A Taxonomy of Middleware: How User Tools Can Improve Social Media

Social media platforms have proven to provide users with a space to connect that is non-existent in the physical world. However, users struggle with mental health concerns, disinformation, and more. Social media’s issues have a wide range of implications for these individuals and society as a whole.

Middleware—third-party services that give users more control over their use of a platform—can counter platform control and provide a powerful complement to necessary market regulation.

Middleware existed on the Internet long before social media came into play and has proven capable of reforming problems in communication software. One example is Pretty Good Privacy (PGP), an encryption program released in 1991 that adds an extra layer of security to online communication, primarily email. Eventually, PGP became the de facto standard for email security because it was free and designed so that users who had never met could send encrypted messages to each other whether or not their email providers wanted to support encryption.

This paper provides a taxonomy of existing middleware in the social media space, organized into five categories with three examples per category. The goal of this taxonomy is to help lawmakers and researchers better understand social media middleware as it currently stands. Here are the categories of the taxonomy, their definitions, and examples:

Category	Definition	Example
Account Aggregation Tools	Multiple social media feeds aggregated into a single feed	Grayjay
Commercial Analytics Tools	Assists businesses in social media advertising	Sprinklr Social
Personal Analytics Tools	Provides users with real-time analytics on their profile not provided by the platform	Reports+
Privacy Tools	Gives users control over their data by subverting a platform’s ability to control, collect, and sell users’ data	Block Party
User Experience Tools	Allows users to consume content on a platform in a manner distinct from the platform to improve a user’s experience on the platform.	Tournesol

With this landscape established, it is clear that middleware already has a significant online presence and the potential to be a legal and functional method for social media reform. A robust middleware marketplace would give users greater agency on social media rather than treating them purely as subjects to be protected by legislators or exploited by platforms. Such a marketplace can also complement regulation with commercial tools that seek similar ends.

My enthusiasm for middleware has limits. For instance, Grayjay (Account Aggregation example) uses content creators’ content without their consent. Reports+ (Personal Analytics example) allows users to access ethically questionable analytics, such as who has blocked you and who is viewing your profile.

I tried to determine the right boundaries for these tools by conducting a stakeholder analysis of some of the examples after the taxonomy. This analysis is rough and meant to serve as a model for a more vigorous examination. Finally, after the stakeholder analysis, I proposed some potential regulations to enable more middleware products to exist within these boundaries.

The takeaway from this paper is that creating a healthy and robust social media middleware landscape will be difficult; helpful programs occupy the same internet as harmful ones. Still, many products have shown potential to develop this space, which should be enough to utilize middleware as one tool to help solve the social media problem.

Link to Paper

Access the full PDF of Isaac Brickman’s “A Taxonomy of Middleware: How User Tools Can Improve Social Media”: http://bit.ly/4cIAol8

Full Text of Paper

What follows is the entire text of this paper. Footnotes could not be included in this post due to formatting. Please reference the above linked PDF of the paper.

Introduction

Big tech companies such as Meta, TikTok, and YouTube have an oligopoly over the social media landscape due to their disproportionate influence and concentration of power. This oligopoly allows a small number of social media platforms to exert large amounts of control over the public sphere and an end user’s experience, creating a space where user and platform interests often come into conflict.

The conflict between user and platform interests raises several concerns. The concentration of the creation and distribution of information on very large online platforms (VLOPs) gives those platforms tremendous power over the digital public sphere. Even though these platforms have facilitated modern democratic movements such as the Arab Spring by providing access to information, there is a bipartisan concern for VLOP’s in America gatekeeping access to information, especially in today’s hyperpartisan climate. Generally, conservatives contend that their content is unfairly restricted, while liberals assert that hate speech is too unmoderated.

Alongside concern for platform’s effects on democracy is a range of worries connecting social media and mental health, particularly the mental health of young adults. Other advocates point out that current models for social media collect vast amounts of data from users in order to profit from targeted advertising, compromising user privacy.

There is a wealth of research dealing with social media concerns. For example, New York University professor Johnathan Haidt argues in The Anxious Generation that social media leads to widespread mental health issues, particularly anxiety, depression, and addiction in teenagers. Others find Haidt’s conclusion are based on faulty research, but acknowledge social media harms certain teenagers. Regardless, the publicized research Haid engages in has led to numerous proposed and enacted laws and regulation. For instance, the surgeon general under Biden proposed warning labels for social media due to the supposed mental health effects on teenagers. Numerous states have also enacted laws or have pending laws geared towards minors, such as banning platforms for users under a certain age, requiring parental consent to create accounts, changing content for minors, and requiring age verification measures.

These regulations could impede free speech concerns. For instance, the ACLU addressed Florida’s 2024 HB 3 social media law, which bans users under fourteen from using social media while requiring age verification by writing, “HB 3 raises significant constitutional concerns because it intrudes on the free speech rights of adults and minors. The internet, including social media platforms, contains vast amounts of constitutionally protected speech for both adults and minors. This government censorship law places unnecessary barriers between adult users and their constitutional right to speak online.” It is worth asking what ‘barriers’ are needed to exercise one’s First Amendment rights on platforms.

The takeaway here is most of the solutions I laid out seek to control users through regulation. Yet there are also ways to reform social media that seek to give users control over their social media experience.

Middleware, the use of a third-party service to give users more control over their use of a platform, is one such way. It can be a viable alternative or a powerful complement to regulation that leverages the market. Middleware existed on the Internet long before social media came into play and has proven capable of reforming problems in communication software. One example is Pretty Good Privacy (PGP), an encryption program released in 1991 that adds an extra layer of security to online communication, primarily email. Eventually, PGP became the de facto standard for email security because it was free and designed so that users who had never met could send encrypted messages to each other whether or not their email providers wanted to support encryption.

According to Stanford professor Francis Fukuyama, middleware, in the context of social media, are “software products that can be appended to the major internet platforms.” Fukuyama’s definition is the first to define the environment and is helpful, but incomplete. Middleware includes products that subtract, not append, functionality. And a definition that focuses on adding software to an existing relationship opens a can of worms – is a VPN used to evade internet censorship considered middleware if it changes one’s interaction with a platform?

Building on Fukuyama’s definition, I propose that middleware should be defined as the following: third-party products that access a social media platform’s data to offer users affordances that differ from a platform’s default affordances. Third-party products refer to commercial and open-source software owned and operated separately from the social media platform the said product accesses. This could include anything from browser plugins to standalone software programs. “Social media platforms” is a similarly broad term in the middleware definition. It means any product where a user can produce content for other user’s consumption and/or consume content that other users produced. Affordances are simply what features a platform offers. For example, the algorithm that shows users content is an affordance. And if a middleware product alters said algorithm, the middleware product offers the user an affordance that differs from the platform’s default affordance. This definition is complicated, and complicated definitions benefit from examples. This paper provides a taxonomy of existing middleware in the social media space, using my definition, with the goal of allowing lawmakers and researchers to understand social media middleware better as it currently stands.

The taxonomy borrows the idea of a “logic” from Rajendra-Nicolucci and Zuckerman’s “Field Guide to Social Media:” a set of business models, technical affordances, and social norms that describe a category of social media. This taxonomy includes five types of tools: account aggregation, commercial analytics, personal analytics, privacy, and user experience. These classifications are meant as broad categorizations from a wealth of middleware products. These categories are not comprehensive, yet the categories will describe the general patterns that differentiate groups of tools in the social media middleware space. Each tool is illustrated with at least three examples of products demonstrating unique affordances within the category. The examples help clarify the details and differences between middleware applications categorized the same way.

Additionally, I follow examples with a table comparing the following nine characteristics of every middleware product in a tool: number of compatible apps, use cases, revenue model, alternative client or helper app, data collected, client or server-side access, type of interoperability, effect on platforms revenue, and change in the content consumed or produced. These features demonstrate the collective affordances and purpose of products in the same grouping to allow comparison between categories, while the examples seek to differentiate applications within the same category.

With this landscape established, it is clear that middleware already has an appreciable online presence and the potential to be a legally and functionally viable method for social media reform. The prospect of creating a marketplace where users have technological self-determination is intriguing; it gives the user agency on social media rather than treating them purely as a subject to be protected by legislators or exploited by platforms. It also offers the possibility of averting overbroad regulation through market channels or complementing regulation with commercial tools that seek similar ends.

Other parts of the internet stack, like browsers and email clients, already provide users with paths towards increased technological self-determination; the Institute for Digital Public Infrastructure Lab, which I am a part of, refers to these parts of the internet stack as “loyal clients.” Loyal clients give users control over how data is shared and how content is organized, allowing them to override decisions made by platform companies. For example, a browser allows a user to block pop-ups and ad trackers despite the preferences of companies that sell ads. Middleware can be a form of a loyal client, as it potentially protects a user’s online interest at the expense of other actors. However, as we’ll see in the taxonomy, this is not always the case.

While government-led reform paths should be part of the conversation, so should middleware. Middleware personalizes the end user’s experience based on the principles of competition rather than regulation. I believe that lawmakers should introduce legislation to allow for a more sprawling and accessible middleware marketplace, where middleware providers would feel legally safe offering adversarially interoperable programs with major platforms, specifically through API access.

My enthusiasm for middleware has limits – I do not advocate for an anarchic environment where any product can interoperate with any platform regardless of consent. The following question can be raised to investigate why: if a third-party product finds a market for a feature not offered by a platform, why would the platform not just adopt the feature themselves? There are good reasons for a platform not to adopt some of these products’ functionalities. These reasons include directly harming an influencers revenue stream by taking away advertisements, allowing third party advertisers access to too data not owned by the middleware user, and providing sensitive analytics to middleware users for users they interact with. These problems will be addressed in the legal implications of existing middleware section after the taxonomy.

Characteristics

Number of compatible apps. The number of social media platforms a product is interoperable with.

Use cases. There are two types of use cases: Personal and Commercial. Personal refers to a user utilizing middleware to improve their experience on a platform. Commercial refers to a company or influencer utilizing middleware as a marketing tool for platforms.

Revenue model. How and if a middleware product makes money.

Alternative client or helper app. An alternative client is a middleware product that serves as a different application used to access a platform. Social media platforms tend to dislike alternative clients because advertising revenue depends on a user seeing the ad on the platform. A helper app is a middleware product that does not serve as an alternative client—instead, it works with an existing client to give it additional functionality.

Data collected. A platform can collect three types of data: Usage, Public, and Private. Usage data is the data a middleware product collects on its users. Public data is publicly available data a middleware product collects, meaning access to the data is not gatekept by the platform. Private data is data a middleware product collects that is not owned by the user or publicly available; this could mean something like a DM a user with a private account sent–private data is gatekept by a platform. These types of data collection are very simplified and will be elaborated on for each example. The categories of data collection also include a middleware tool allowing third parties to collect data on their software.

Client-side or server-side access. Client-side access means access to the data which only rewrites what a user sees. This type of access usually comes in the form of a browser extension, as what is changed is the interface such as the CSS (accessible on a browser), not the internal data only accessible on the server. Server-side access is access to data that sends information to the server and potentially asks the server to change behavior. These types of products normally are contingent on API access, since API access is necessary to obtain information only available on a server, such as what content does X user typically engage with.

Type of interoperability. Interoperability means a product can interact with another product’s files or data—Keynote, the presentation software from Apple, can open, edit, and write PowerPoint files. There are three types of interoperability: Adversarial, Indifferent, and Cooperative. Adversarial interoperability refers to a middleware product that operates with a social media platform without the platform’s express permission. One might posit that if a platform does not want a middleware product to operate with them, the platform probably has a good reason for it, as a platform would claim that they have every right to have authority over how its intellectual property is utilized. Yet some of the most prevalent software companies have achieved scale through adversarial interoperability.

Apple achieved success in the early 2000s partly by interoperating with Microsoft applications such as Word, Excel, and PowerPoint on their own product suite, iWork, to allow for cross-platform workflow. However, Microsoft tried to continually update its products, so they were not compatible with Apple’s. Apple’s engineers managed to keep overcoming Microsoft’s barrier, and eventually, Microsoft had to accept compatibility because Microsoft’s users complained that they could not have a cross-platform workflow.

In the mid-2000s. Facebook (now Meta) also enabled their users to message MySpace users, allowing them to overcome the network effects problem that gave MySpace market dominance. Now, users would rather have Meta than MySpace—evident in Meta’s rapid growth and MySpace’s concurrent decline. Therefore, adversarial interoperability tends to disrupt a dominant player but can lead to innovation, giving users a higher-quality product. The examples of Apple and Meta do not necessarily mean adversarial interoperability is good, but they do indicate that a middleware product should not be dismissed out of hand because it is adversarially interoperable with a social media platform.

Indifferent interoperability refers to a middleware application that a social media platform is not taking any action towards. Cooperative interoperability is a middleware product that a social media platform actively supports. Indifferent and cooperative interoperability are relevant because they indicate a platform is at the least okay with the middleware program. If middleware is to be utilized as a solution to social media reform, indifferent and cooperative interoperability illustrate a platform’s negotiating position towards what should be allowed.

Effect on platform’s revenue? A positive effect on revenue would entail a monetary benefit for a platform, while a negative effect would have the opposite effect. A middleware product with a negative effect would intuitively necessitate adversarial interoperability. It often does, but this is not necessarily the case.

Let’s use Axel Springer vs. Adblock Plus as an example to elaborate. Axel Springer vs. Adblock Plus is an ongoing case that pertains to the legality of data access based in Europe. Axel Springer is a Berlin-based multinational mass and online media company that publishes online advertisements. Axel Springer brought a case against Eyeo, the provider of the Adblock Plus ad-blocking tool, for providing a feature that blocks ads, including Axel Springer’s. The company claims that Adblock Plus violated the statute prohibiting “an unauthorized modification of a computer program” to which Axel Springer has exclusive use. A lower court dismissed Axel Springer’s claim because an ad-blocking feature does not actually change the substance of the program; it merely changes the “display of the HTML document as a temporary intermediate result during the execution of the website programming.” The lower court also emphasized how the installation of the add-on was the user’s decision. Axel Springer vs. Adblock Plus is relevant because the case illustrates how novel disputes revolving around data access are being resolved and serves as a guide for how courts see middleware that accesses a platform’s data. Also, the case shows that undercutting a platform’s revenue often depends on server-side access to be adversarially interoperable because laws built around third parties using data make liability contingent on access to internal data. Therefore, if the adblocker accessed Axel Springers server-side data, it is reasonable to think the case could have gone differently.

Change in the content consumed or produced? If a middleware product changes the content consumed, it means the application is designed to expose its user to content that looks distinct or is different from the original content on a platform. If the middleware application changes the content produced, it is meant to alter the content that a user produces.

Now, I will examine several middleware platforms, considering these different affordances, and cluster them into five conceptual buckets, or ‘tools.’

Account Aggregation Tools

People occupy many different roles. For example, I am a son, friend, student, mediocre recreational basketball player, occasional hiker, and tutor. Fulfilling all of these roles can be demanding. If I am a student at 2:00, I will have to shift my mindset to teach elementary schoolers at 3:00, and I will not treat elementary schoolers like my friends at a party at 9:00.

Similarly, I also have different roles online. I check Instagram to keep up with my friends and catch up on DMs. I go to YouTube to enjoy the content of my favorite creators and use Snapchat to send people useless selfies of myself. Account aggregation tools seek to reconcile the inconvenience of juggling different selves online by using middleware to combine different online selves onto a single feed. Because content from multiple social media platforms is aggregated into a single feed, all middleware products that are account aggregation tools are alternative clients that change the content consumed. It is essential to note some examples of middleware products offer aggregation as a feature but are not considered account aggregators in my taxonomy because the aggregation is only supplemental to the application’s primary function.

Account Aggregation tools are not new. Really Simple Syndication (RSS), developed in 1999, allowed users to aggregate content from different websites into a single data feed. RSS syndication was particularly important for websites whose content frequently changed, like weblogs or news sites, and popular RSS aggregators – newsreaders – became popular ways to keep up on news from bloggers and professional journalists. Now, many RSS products enable users to aggregate social media feeds. Certain RSS providers also allow their customers to produce content. For example, users can upload their podcast to an RSS feed and distribute it across all the major podcast platforms. Indeed, podcasting at its heart is as simple as distributing an audio file and metadata via RSS.

Messenger for WhatsApp Duo Web

A less complex account aggregation tool is Messenger for WhatsApp Duo Web (https://apps.apple.com/us/app/messenger-duo-for-whatsapp/id1312424138), developed by BEST SOCIAL APPS LTD. The app allows a WhatsApp user to aggregate multiple WhatsApp accounts onto a single feed. Aggregating WhatsApp accounts can be helpful in many circumstances, from legitimate ones like having a separate business and personal account to more nefarious ones, like a ‘Mrs. Doubtfire situation’ or an online scammer.

Messenger for WhatsApp Duo Web is much different than Grayjay. For one, Messenger for WhatsApp Duo Web is not adversarially interoperable with WhatsApp. Instead, I characterize Messenger for WhatsApp Duo Web as indifferently interoperable, meaning WhatsApp does not actively fight or promote its existence. Messenger for WhatsApp Duo Web is also only compatible with a single app, WhatsApp, unlike Grayjay, which is compatible with eight different platforms.

Messenger for WhatsApp Duo Web’s premium version allows users to link their business and personal accounts (requiring a subscription). Therefore, it is designed for both commercial and personal use. In general, business uses of social media middleware will involve at least some content creation, as social media is generally used to communicate with customers or employees through these channels.

According to Messenger for WhatsApp Duo Web’s privacy policy, they collect usage data and share such information with their ‘affiliates,’ including their parent company and business partners, neither of which I could find. This raises questions about privacy and possible data breaches. The middleware product also states, “when You share personal information or otherwise interact in the public areas with other users, such information may be viewed by all users and may be publicly distributed outside.” This statement implies that Messenger for WhatsApp Duo Web also collects public and private data, as the product collects a user’s interactions. Messenger for WhatsApp Duo Web also has client-side access – it is not merging accounts on the server or telling WhatsApp that two or more accounts are associated with each other – instead, it is simply combining their data streams in a custom client.

Grayjay

Grayjay (https://grayjay.app/), developed by FUTO, aggregates video from multiple video hosting platforms while advocating a very specific political agenda regarding tech. Grayjay’s creators believe that YouTube has a monopoly over video-sharing platforms because of network effects. Yet some video creators use other platforms, such as Odysee and Patreon, which are supported via donations to the creator rather than through monetizing ads. YouTube users might be fans of a creator’s content on YouTube but know that the same creator posts other content on Patreon. Yet this fan might be unwilling to download and use Patreon to watch a single creator. Grayjay solves this problem by giving users the option of aggregating eight video-sharing platforms into a single feed, as their philosophy is that users should follow creators, not platforms.

Grayjay’s is designed for personal consumption, not commercial use. Furthermore, unlike YouTube, watching a video on Grayjay means no advertisements. Instead, Grayjay makes its revenue from a one-time payment. According to their privacy policy, the application has a user’s locally stored data, meaning the data that pertains to one’s use of the app, but they do not access it. Grayjay claims to only access a user’s country of origin and postal code for tax purposes on a purchase but does not sell such data to third parties. This means that Grayjay only collects usage data. The program also accesses client-side information – on the platforms it is interoperable with such as subscribers, likes, and the videos themselves.

Grayjay takes YouTube users off of YouTube by allowing YouTube’s users to watch creators from multiple social media platforms in a single feed, which means Grayjay has a negative effect on YouTube’s revenue. Perhaps understandably, YouTube sent a cease-and-desist to GrayJay’s creator, Louis Rossmann, which means I can consider Grayjay as adversarially interoperable.

The cease and desist accused Rossmann of violating YouTube’s Application Programming Interface (API) Service’s Terms of Service (ToS) and developer policies. The API Service’s ToS policies require Rossmann to clearly display a link to YouTube’s ToS and state in Grayjay’s terms of use that YouTube’s ToS and privacy policy binds users of Grayjay. YouTube also accuses Rossmann of illegally using YouTube’s content on Grayjay and demands Rossmann delete all YouTube content from the app.

Rossmann responded in a video explaining that his product never agreed to YouTube API ToS or developer policies because Grayjay or its end users do not access YouTube’s API services. He subsequently claimed that YouTube seeks a monopoly by bullying smaller video-sharing platforms like Grayjay into submission. It’s fair to note that a court may not appreciate Rossman’s legal reasoning here, as the application is still using YouTube’s content without permission.

While eliminating YouTube ads may be a selling feature for some users, Grayjay has additional and unique functionality. Grayjay helps users find creators they follow more easily than other platforms, showing users a feed that only contains creators the user follows, not a feed that recommends content from any creator based on consumption patterns. Grayjay’s algorithm gives users and creators a more direct relationship, where users will not have to seek out content from a specific creator or hope that the creator shows up in their feed. Grayjay also allows users to create custom playlists and organize videos based on their preferences. In this sense, Grayjay delivers something users have often asked for – a social media feed governed entirely by preferences and not by opaque algorithms which studies have shown “often incentivize conflict actors toward more divisive and potentially violence-inducing speech, while also facilitating mass harassment and manipulation.”

With all that said, Grayjay is an imperfect David to root for to take down the YouTube Goliath. Creators will not receive ad revenue when users watch their videos on Grayjay unless they directly support a creator through Patreon or a similar platform. Furthermore, Grayjay would do this without a creator’s permission, essentially forcing creators into a new financial bargain (it’s worth noting creators were already forced into an unfortunate financial bargain by platforms like YouTube.) Arguably, the product should not exist, given its potential to materially harm influencers whose livelihood depends on these platforms while only moderately helping users. At the same time, Grayjay has some good features that should be considered. Namely, Grayjay helps subvert YouTube’s video-sharing monopoly by overcoming the network effects problem, giving users more choice and smaller video-sharing platforms a chance to gain scale. And, importantly for the purposes of this paper, Grayjay shows just how complicated it can be to evaluate the impacts of and ethical decisions behind middleware projects.

Gobo

Gobo (https://gobo.social/), developed by the Initiative for Digital Public Infrastructure Lab at the University of Massachusetts Amherst, is an experimental middleware program currently only open to a group of beta testers. Gobo aggregates BlueSky, Mastodon, and Reddit into a single feed and allows posting to those three and LinkedIn. These platforms Gobo are interoperable with offer diverse features, which is by design – Gobo’s creators wanted to demonstrate that social networks designed for different audiences could be made interoperable. Gobo’s affordances are as follows:

Read and post to multiple social networks from one open source client.
Pick and choose between algorithms for filtering and sorting the posts from all these networks.
Design (and potentially share) different algorithms for filtering and sorting.
Use third-party services to assist in filtering and sorting.
Audit the performance of these different algorithms as well as the third party services.

Gobo allows users to contribute to the design of the app as an open-source client and encourages the use of third-party algorithms, making Gobo both an aggregator and a middleman for middleware offerings. The product is meant for personal use, is free, collects usage data purely for research purposes, has client-side access, and is indifferently interoperable with the social media platforms it uses. However, it would negatively affect a platform’s revenue because it diverts traffic, reducing ad views.

Gobo can be helpful when people might not want to switch between platforms to reach a broad audience on X (Twitter), converse in a server that moderates harassment on Mastodon, and discuss their favorite Adam Sandler movies in an Adam Sandler subreddit. Gobo also illustrates the potential for a healthy social media marketplace because it is not profit-oriented, allowing Gobo to adopt features that are best for the customer without considering revenue. Placing the customer first is fundamental to a potential environment where middleware is regularly used to improve a user’s social media experience. As social media has demonstrated through an extractive attention economy, profit-motivated market innovation will not necessarily improve a customer’s experience. However, it should be noted Gobo does not charge purely because it is a research experiment supported by foundation grants.

RSS.app

RSS applications are by far the most common account aggregation tools. RSS.app (https://rss.app/), developed by FeedsApp Inc., works by inviting a user to enter a keyword or hashtag and generating a feed related to the keyword or hashtag with filters that further specify the content A user can ‘bundle’ feeds from multiple platforms, allowing it to aggregate across numerous platforms. For instance, if I love dogs (which I do), I can use #cutedogs to find pictures related to cute dogs from Instagram, Facebook, and Reddit. RSS.app is designed for personal use because it displays a feed that a user consumes. The social media feeds are only available with a subscription. The app collects personally identifiable information (PII) (name, email address, phone number, etc.) and sells the PII to third parties. PII should be a principal part of the middleware conversation because PII often includes sensitive information users do not want third parties to know about. RSS also collects public and private data. I know this because they say in their privacy policy, “We may allow selected third parties to use tracking technology on the Sites, which will enable them to collect data about how you interact with the Sites over time.” By allowing third parties to track how one interacts with the site, they are, by default, tracking interactions with other users. Like PII, this is another crucial concept in middleware; it raises the question of how applications should be able to use data from users who do not use the tool.

RSS.app also has client-side access. Lastly, accessing a platform’s RSS feed necessitates cooperative interoperability because the social media platform must provide a feed for the RSS to access. However, since the RSS diverts traffic, it negatively affects a platform’s revenue. It is then speculative why RSS is cooperatively interoperable with platforms, but it might be because they are “grandfathered in.”

Characteristics of Account Aggregation Tools

Considering the characteristics offered in the account aggregation tool examples, let’s create a table comparing the characteristics. Features highlighted in yellow are the same across all the applications for this tool.

Account Aggregation Tools Characteristics
Number of compatible apps	Varies
Use cases	Often personal, occasionally commercial
Revenue model	Varies, including undercutting existing business models and non-profit, arguably unsustainable models
Alternative client or helper app	Alternative client
Data collected	At least usage data
Client-side or server-side access	Client-side access
Type of interoperability	Adversarial, indifferent, or cooperative
Effect on platform’s revenue	Varies
Change in the content consumed/produced?	Consumed

Number of compatible apps. One might think aggregators must be compatible with multiple social media platforms. While compatibility with numerous social media platforms seems intuitive, Messenger for WhatsApp DuoWeb demonstrates that aggregators for a single platform can be useful.

Use cases. Middleware products that are account aggregation tools must be designed at least partly for personal use because aggregators allow users to consume content from multiple sources in a single feed.

Revenue model. Aggregators offer a variety of payment models.

Alternative client or helper app. All account aggregation applications are alternative clients because no platform allows multiple user profiles to be displayed in a single feed, making an alternative client necessary.

Data collected. Data access is varied. However, all these products access at least some form of usage data.

Client-side or server-side access. All account aggregation middleware has client-side access because these applications can only change the interface or content a user is exposed to.

Type of interoperability. Aggregators besides Grayjay are not adversarially interoperable, and RSS feeds are cooperatively interoperable with platforms.

Effect platform’s revenue? Most aggregators have a negative effect on the platform’s revenue. But WhatsApp does not advertise, but makes money by partnering with businesses to communicate with users. Therefore, WhatsApp Duo Web, being an alternative client, is irrelevant to WhatsApp revenue.

Change in the content consumed or produced? Account aggregation tools change how content is consumed because users are exposed to different content than they would be without the tool.

Commercial Analytics Tools

Because social media is so widespread, businesses are taking advantage, using platforms to reach an extensive audience. This has spawned a market of middleware providers offering tools to improve the convenience of content production, user engagement, and drive revenue. Social media middleware providers that are commercial analytics tools operate similarly to account aggregation tools, as a major purpose of these platforms is to interact with multiple platforms simultaneously. The tool differs because account aggregation tools primarily allow users to consume content from various platforms in the same feed. In contrast, commercial analytics tools mainly enable customers to produce content on multiple platforms. Because of this, all commercial analytics tools products change the way content is produced

Most commercial analytics tools have features like scheduling posts. This can work well for customers like influencers who might not want to manually post that their new YouTube video has come out on numerous platforms. Commercial analytics tools generally have API access to platforms to provide analytics on a user’s post performance. Commercial analytics tools geared toward large companies provide advanced analysis on content engagement and access to an extensive repository of public data that offers insights into current trends. Finally, commercial analytics tools must be used almost exclusively for businesses and interoperate cooperatively with platforms (which helps separate them from personal analytics tools).

Sprinklr is a software-as-a-service company that helps businesses with their marketing and customer service needs. It offers four product suites, including a social media tool, Sprinklr Social (https://www.sprinklr.com/lp/social-media-management-enterprise), which allows its customers to create, manage, and optimize their social media interactions with customers. The program offers many features, including AI-driven analytics, ‘social listening’—real-time access to a large repository of publicly available data that enables clients to unlock trends and long-term insights, and hashtag optimization.

Sprinklr Social interoperates with forty-three social media platforms and is used by brands such as Microsoft, McDonalds, and Honda. It offers a subscription-based revenue model with a thirty-day free trial, with plans starting at $3588 annually. This application and all middleware providers that are commercial analytics tools can be considered helper apps rather than alternative clients. According to their privacy policy, Sprinklr Social only collects publicly available data. This seems to be the norm for all commercial analytics tools that employ analytics to reach a wide user base as well.

Analytics are also a central part of any commercial analytics tool, and analytics need information on how users interact with a business’s content. Because analytics necessitate access to a platform’s internal data, all analytics products have server-side access.

Sprinklr Social, like most commercial analytics tools, is cooperatively interoperable with platforms, since, to conduct analytics on social networks, which is a central feature of commercial analytics tools, the platform must grant the software API access. It is not surprising that commercial analytics tools are the only type of middleware where cooperative interoperability is commonplace, as programs in this category can afford API plans. Because these tools streamline a business’s advertising, platform revenue also increases. Therefore, platforms hope that API will be used by tools like Sprinklr, which pay platforms for access and process useful data for client analytics.

APIs raise the question of whether API access should be kept behind a paywall. Not all platforms gatekeep access, but the ones that do certainly hurt a diverse set of developers. This includes academics who use large-scale data to conduct social media research, as in 2023, X started charging $42,000 per month for API access, effectively pricing out researchers. Because platforms place API access behind a paywall, API has become a double-edged sword for developers. On one end are the developers who can pay. On the other are those that cannot. Effectively, platforms can dictate the affordances developers can offer based on their profitability, not their benefit for the user.

Buffer

Sprinklr Social is trying to gain McDonalds as a client, but Buffer (https://buffer.com/), which is compatible with eight platforms and only offers analytics for Facebook, Instagram, X, and LinkedIn, is geared more toward social media managers and small businesses. Buffer’s advantages include a more straightforward interface and low prices, with a free version for individuals and various plans from six to one hundred dollars a month.

Buffer prides itself on providing transparency, as their “About Us” section categorizes itself as an open company. For them, this means posting their finances, such as monthly recurring revenue, net income, and bank balance. They also show how many active users they have and team member salaries, along with where every employee is located (the company is remote) and their hobbies. Considering the number of scandals VLOPs have gotten into for lack of transparency, Buffer’s practices could be a good starting point for how middleware applications approach transparency in a future social media middleware landscape.

Repost+ for Instagram

Repost+ for Instagram (https://apps.apple.com/us/app/repost-for-instagram/id1477820240), developed by Livintis W.L.L., advertises its primary functionality as allowing a user to repost another user’s story, reel, photo, and video, while giving credit through a watermark to the original poster. Therefore, this product would increase content production, thereby increasing engagement and traffic which supports Instagram’s revenue. Repost+ is also much different from Sprinklr, Buffer, and most other applications that are commercial analytics tools. It is only compatible with Instagram, does not offer analytics, and is indifferently interoperable with Instagram. That being said, it offers many similar features, including post-scheduling, AI hashtags, and caption templates. Because Repost+ uses a separate app to repost content, it does not simply access Instagram’s interface, giving this product server-side access.

The app does require a subscription, but it is relatively cheap at only $2.99/week. The app also only collects certain personal data that can be turned off in its settings; they do not sell this data to third parties. Due to its price, lack of features, and compatibility with one platform, the product is generally used by small businesses or influencers, who might not have the scale that requires analytics or budget to hire a social media manager who can post on numerous platforms.

Characteristics of Commercial Analytics Tools

Middleware applications that are commercial analytics tools are much more widespread than aggregators. While there is no exact count, there are easily hundreds of such products. Despite this, commercial analytics tools must contain a much more homogenous set of characteristics than aggregators.

Commercial Analytics Tools Characteristics
Number of compatible apps	Varies – from 1 to 43 in our examples
Use cases	Commercial
Revenue model	Subscription-based, although some have a free version
Alternative client or helper app	Helper app
Data collected	Varies
Client or server-side access	Server-side access
Type of interoperability	Indifferent or cooperative
Effect on platform’s revenue?	Positive
Change in the content consumed/produced?	Produced

Number of compatible apps. Commercial analytics tools are usually compatible with multiple applications because brands like to utilize numerous platforms, but the exact number differs.

Use cases. An app is not a commercial analytics tool unless its usage is primarily commercial.

Revenue model. The affordances offered, and the clientele who use this tool make a non-subscription-based revenue model unlikely.

Alternative client or helper app. These products are helper apps, as alternative clients are primarily meant for content consumption.

Data collected. Most collect customer, public, and private data needed to display analytics, but Repost+ for Instagram does not.

Type of interoperability. Commercial analytics tools software is cooperatively interoperable due to API and RSS access.

Does it undermine a platform’s revenue? These products either access an API, increase content production, or both. Therefore, all these products increase a platform’s revenue.

Does it change the content consumed or produced? It changes the content produced because features in all commercial analytics tools are meant to assist companies or influencers in producing content.

Personal Analytics Tools

Want to know who stopped following you on social media? What post cost you the most followers? You might be in the market for a personal analytics tool.

Most middleware providers that are personal analytics tools advertise themselves as aligned with commercial analytics tools. At first glance, they seem this way, providing users with real-time analytics on the state of their profile. However, with a closer look, it can be argued that personal analytics tools are less about analyzing the state of one’s business on social media and more about the social standing of one’s personal life.

For example, middleware that are personal analytics tools all contain roughly the same features, including: ‘secret admirers,’ ‘ghost followers,’ ‘lost followers,’ and ‘who blocked me.’ These features seem geared toward finding out what other users think about you or your brand. While there are legitimate business cases for determining why content is chasing away followers, these programs generally present themselves less professionally than middleware which are commercial analytics tools.

Based on numerous reviews on the app store and Reddit, it is reasonable to presume most of these applications can operate in ways that can harm the user, like getting users logged out of or even banned from Instagram when logging into the platform via the middleware software to sync their profile and view analytics.

These same reviews attest to the fact that many of these programs do not function, but almost all that do likely send bots through Instagram’s Insights API that performs actions on behalf of a user to collect information about their page, including social listening metrics. Therefore, these tools have server-side access. Instagram also most likely tracks these bots by adopting rate-limiting protocols, which cap the number of actions a user can take in a specific timeframe. Because Instagram is trying to undermine the middleware function, the personal analytics products that are interoperable with Instagram are adversarial. Also, Instagram might be opting to use a protocol instead of shutting these tools down manually because it is easier for them not to go after each app. These products don’t really affect a platform’s revenue – further explaining why Instagram is not especially aggressive about shutting them down – or change content; they only offer insights into other users. Finally, tools in this category are important to analyze because it is worth looking into the ethics of a business model built on access to customer data that a platform does not provide.

Unfollow for X aka Twitter

Unfollow for X aka Twitter (https://apps.apple.com/us/app/unfollow-for-x-aka-twitter/id369327398), developed by Spicy Apps, is the only personal analytics tool for a platform other than Instagram. Because X is more open source than Instagram is not, one might initially expect X to be the platform most personal analytics tools are compatible with, not Instagram. This is probably not the case because Instagram provides more of a ‘friends and family’ functionality than X.

Unfollow for X aka Twitter is also much more stripped-down than other personal analytics tools. The app allows one to unfollow inactive users, unfollow those who do not follow back, and find new unfollowers/followers. It does not contain features that can be viewed as a violation of other user’s privacy, such as ‘secret admirers’ or ‘who blocked me.’ Like all other personal analytics tools, it offers a free version with a subscription option. Unfollow for X aka Twitter also accesses usage data in the form of PII, which it sells to third parties. The app is also cooperatively interoperable with X as it uses its API services, and X is not actively trying to impede the products function.

Panda Spy

Panda Spy (https://apps.apple.com/us/app/panda-spy-reports-from-insta/id1107522062), developed by Asian Turbo Limited, is like many other apps that are personal analytics tools in the sense that they are seen by many users as dysfunctional, and have the same functionality as dozens of other tools on the market. For example, one review claims that after the free trial period ends, the subscription is impossible to cancel because to do so, a customer must go to the website, which is under construction. Other applications that are personal analytics tools have cancelable subscriptions. Still, most follow the same revenue model: free features (usually ones users can find without the software, like those following them) and a subscription.

Reviews say Panda Spy does not work for Facebook, X, or Instagram. I could not find evidence contrary to these claims. Because there is no evidence that the application works, it is reasonable to categorize the program as compatible with zero platforms.

Like all personal analytics tools, the product is a helper app that may or may not collect public and private data because I do not know if the software works… although Panda Spy most likely gathers PII, and they do say they do not sell such information to third parties without consent.

Reports+

Reports+ (https://apps.apple.com/us/app/reports-followers-tracker-ig/id1481469888), developed by GopGop Apps, is the most mainstream personal analytics tool. The product, which is compatible with Instagram, has 339,000 ratings on the Apple Store and is number thirty-three on the social networking charts at the time of writing this. The application offers various features, including ones most of these tools have, like ‘secret admirers’ and ‘ghost followers.’ This product also has unique features, such as informing the user when an account gets reactivated or disabled, and a profile discover feature that allows one to locate any account with autocomplete suggestions.

However, potential users should be aware of the popularity of this product. It is an app that gets users logged out of Instagram upon use, which is the only platform the software is compatible with. Many reviewers also seem worried that the app is trying to collect the user’s login credentials, which is possible. It is on their privacy policy that the app collects PII and sells it to third parties.

Characteristics of Personal Analytics Tools

Like commercial analytics tools, personal analytics tools contain many middleware applications and offer predictable affordances. Unfollow for X aka Twitter is the only app that varies significantly from the rest.

Personal Analytics Tools Characteristics
Number of compatible apps	Instagram, and X. Or none, as some products don’t work
Use cases	Personal
Revenue model	Subscription-based with free features
Alternative client or helper app	Helper app
Data collected	At least PI, which is sold to third parties
Client or server-side access	Server-side access
Type of interoperability	Varies
Effect on platform’s revenue?	None
Change in the content consumed/produced?	No

Number of compatible apps. All personal analytics tools are compatible with zero or one app, depending on whether it works.

Use cases. These applications all seem to function primarily for personal use, even though many advertise themselves for commercial purposes. This is because these tools mostly contain features such as ‘secret admirers,’ ‘ghost followers,’ or ‘who blocked me,’ which are likely not the names of features a business would like use. This functionality is further proven by the sexuality some middleware products use when displaying features in the Apple Store preview, appealing to young users rather than brands.

Revenue model. Every personal analytics tool I could find is subscription-based with accessible features, which is not a prerequisite for personal analytics tools.

Alternative client or helper app. These programs are helper apps and are not meant to serve as an alternative POS.

Data collected. While a personal analytics tool might not work, it will likely collect at least PII and sell it to third parties. It could also collect public and private data since providing login information is a condition for use.

Type of interoperability. Tools that have API access seem to be at least partially adversarial with the platform, while other tools, such as Unfollow for X aka Twitter, are not.

Effect on platform’s revenue? These products do not really have an effect on a platform’s revenue. One can argue that seeing who is interacting with your profile and how makes one more interested in the platform. However, one can also posit that seeing who unfollowed you would make you unfollow them, thereby resulting in fewer users to engage with. Either way, both arguments seem spurious.

Change in the content consumed or produced? These products do not change the content consumed or produced.

Privacy Tools

Big tech social media companies are infamous for violating user’s privacy. From Meta’s Cambridge Analytica scandal, Meta being fined $1.3 billion by the E.U. for transferring European user’s data to the U.S., to a federal ban on TikTok (pending a sale) because of the Chinese government collecting user’s data, privacy scandals have been persistently making national headlines. Users want a way to protect their privacy on social media platforms without having to get off the platform itself. This is where privacy tools come into play.

Middleware privacy tools seek to give users control over their data by subverting a platform’s ability to control, collect, and sell users’ data by blocking content trackers, advertising, or erasing data. Privacy tools can be helpful in many situations, whether wanting to turn off invasive advertisements or delete all the negative posts about Taylor Swift. All applications that are privacy tools change the content consumed.

Redact.dev

Redact.dev (https://redact.dev/) is the only middleware program I could find that allows users to erase data from a platform. This app is compatible with twenty-four different platforms, but not with TikTok, Instagram, and Snapchat. Although a business can use Redact.dev, its purpose is personal. The software wants to help users delete their history, and companies are less likely to need to delete an embarrassing old tweet than a person would be. For example, a person who made some questionable tweets in highschool, and does not want to delete all of them, might want to use Redact.dev before searching for a job. The application is also subscription-based, with free features that allow users to delete a limited amount of content. The product is also a helper app that collects PII and sells it to third parties. The software is indifferently interoperable with all compatible platforms and does not undermine their revenue.

Despite Redact.dev being indifferently interoperable with platforms, middleware that seeks to delete data could have the potential for legal liability, as bad actors can use these programs to erase evidence. The software’s mission also might occupy a moral gray area beyond the law. For example, what if political candidates use this service to delete questionable tweets about their views on certain marginalized groups? While erasing a person’s past mistakes could be a valiant goal, these negative implications deserve to be contended with.

Cyd

Cyd (https://cyd.social/), developed by Lockdown Systems LLC, is an open source software that offers essentially the same affordances as Redact.dev for X users. Basically, a user will access an application that acts as a separate browser to delete tweets and reweets for free with a premium plan offering the ability to delete likes, DMs, and unfollow everyone. The product accesses some PII not associated with a users X account, but does not sell that information to third-parties. Cyd also intends to be interoperable with other VLOPs such as Facebook as well. While Cyd has the same functionality of Redact.dev, what separates Cyd is that it can transfer Tweets from X to Bluesky (a federated platform that will be covered in detail later). Therefore, Cyd holds the unique title in this taxonomy of moving content between platforms. I would consider this functionality to be in the “User Experience Tools” bucket, however, Cyds website advertises itself as primarily focused with deleting data, so I decided to put this program here. Lastly, the software presumably undermines Xs revenue, mainly by helping users migrate to Bluesky.

Friendly Social Browser (https://apps.apple.com/us/app/friendly-social-browser/id400169658), developed by Friendly Inc., has an entirely different functionality than Redact.dev. This program allows users to take control of their data on Facebook, Messenger, Instagram, X, and LinkedIn by providing their own advertising blocker. It does this by serving as an alternative client and providing its own browser, thereby not having to access the platform’s API. Since Friendly blocks ads, it has a negative effect on their revenue. It is worth noting that this product is not an aggregator because it displays a different feed for each app rather than a single feed for all the platforms it uses.

This software also offers affordances that allow users to personalize their experience, with features such as “ordering feed by most recent,” “filter/highlight feeds by keywords,” and “colorful themes.” These elements mirror the following categorization, user experience tools, but it is more accurate to characterize friendly as a privacy tool, as they advertise themselves on their website as an application that allows users to “take control of your data and privacy.”

While the product claims to allow users to “take control of their data,” the app still accesses usage data anonymously for certain purposes but does not sell it to advertisers. The software is free but offers a Friendly Plus version at ninety-nine cents. This version provides the same features, but some of them, such as “colorful themes,” are “advanced colorful themes” in this version. The developer also offers Friendly for Twitter, a free app offering similar features such as “custom color themes” but also different ones such as “block promoted tweets.” Friendly was threatened with legal action by Facebook in 2020 for “chang[ing] the way Facebook and Instagram look and function” and “impairing [their] intended operation.” However, they have not experienced any legal backlash since, presumably because they do not access Facebook’s API services. This means Friendly is currently indifferently interoperable with platforms.

Anywhere for Facebook

Anywhere for Facebook, developed by Freedom Apps, is essentially a VPN explicitly made for Facebook, Instagram, and Messenger. It works by using a built-in proxy server to hide a user’s IP address. Hiding an IP address is an important way to access information when in countries that censor social platforms (China, Russia, Iran). It is worth noting that there is also an Anywhere for Twitter app, which is essentially the same product. Purchasing Anywhere for Facebook requires a one-time payment of $3.99. The product is an alternative client because, like Friendly Social Browser, Anywhere for Facebook uses an application that provides a separate browser. The application also collects PII and sells it to third parties. Furthermore, the application also has client-side access, is indifferently interoperable with Meta, and helps Meta’s revenue since it gives users access to more content, therefore changing the content consumed.

Block Party

Block Party (http://blockpartyapp.com/) is an application that streamlines the process of configuring 223 privacy settings for users across ten platforms. It does so by acting as a browser extension, running a scan of a platform and creating a checklist of recommendations and automatic changes based on current privacy settings. For instance, on X, location is turned on by default. After Block Party runs a scan of X, it will automatically remove the location from a users profile. After I ran a scan on X, Block Party also recommended me to set up two-factor authentication, and mute notifications from certain accounts, like those I do not follow or have not confirmed their email. The product is designed for commercial use cases as well. For example, organizations use it to identify and remediate overexposure to first-party data. The program offers a yearly subscription with free features, is a helper app, collects PII, which it transfers to third parties with the prior consent of the users, and only for non-marketing situations (request logging, emailing, and error logging). The app has client-side access as it is only a browser extension, is indifferently interoperable, and does not undermine the revenue of any platform it uses.

Also, what is interesting about Block Party, is that its main functionality used to be blocking users who can potentially harass them en masse. For example, if a harasser threatened you in a tweet, Block Party allowed you to block the harasser and everyone who retweeted or liked the post. However, after Musk started charging for API access, Block Party was priced out and lost this blocking functionality that required server-side access. This past version of Block Party will be revisited later in the “What products should platforms be forced to interoperate with.”

Characteristics of Privacy Tools

Privacy tools are complicated. For example, Redact.dev and Friendly Social Browser provide completely different affordances. This makes categorizing these tools difficult, yet there are still some features that all these tools have in common.

Privacy Tools Characteristics
Number of compatible apps	Varies
Use cases	Personal
Revenue model	Varies
Alternative client or helper app	Varies
Data collected	PII
Client or server-side access	Varies
Type of interoperability	Indifferent
Effect on platform’s revenue?	Varies
Change in the content consumed/produced?	Consumed

Number of compatible apps. Most privacy tools are compatible with multiple platforms, but Friendly for Twitter shows an exception to this fact.

Use cases. Almost all users who use a privacy tool purchase the software for personal reasons.

Revenue model. The revenue model varies. Redact.dev is subscription-based. Friendly offers a free app and a similar version for ninety-nine cents. Anywhere for Facebook uses a one-time payment option.

Alternative client or helper app. While Redact.dev is a helper app, Friendly and Anywhere for Facebook’s functionality depends on an alternative point of access to social networks.

Data collected. Privacy tools seem only to access usage data in the form of PII, and some sell it to third parties.

Type of interoperability. Privacy tools are indifferently interoperable with platforms.

Does it undermine a platform’s revenue? This varies. Redact.dev does not undermine a platform’s revenue, Friendly Social does, and Anywhere for Facebook facilitates a platform’s revenue.

User Experience Tools

Suppose I want to change the content I am exposed to on online platforms, or even more importantly, suppose I want to make the text on my feed only contain a medium Playfair Display font. This is where user experience tools come into play. The functionality of user experience tools is based on providing users with affordances to consume content on a platform in a manner distinct from the platform to improve a user’s experience on the platform. User experience tools are almost exclusively designed for users who seek control over the type of content they consume or the interface they engage with the content on, not on what they produce. Because of this, user experience tools are designed for personal use. Giving the user control over the type of content consumed is key, considering content consumption is foundational to how a user interacts with a platform. Changing the content consumed, therefore, can meaningfully impact how social media affects a user, a user’s interactions with other users, and the democratic discourse as a whole.

Narwhal for Reddit

Narwhal for Reddit (https://apps.apple.com/us/app/narwhal-for-reddit/id845422455), developed by Robot Swingset, provides users with a personalized Reddit experience, offering features like a custom interface, fonts, and themes. It is designed for personal use and was a free product, but Reddit started charging for API use, forcing the developers to now charge a subscription. The API also makes Narwhal 2 cooperatively interoperable with Reddit, increasing the platform’s revenue. The program serves as an alternative client, providing a different way to access Reddit. Narwhal 2 collects PII but does not sell it to advertisers unless the user agrees. Interestingly, Reddit hosts many clients with similar functionalities to Narwhal, such as Comet for Redditt, Sink It for Reddit, and Oldr for Reddit.

SocialFocus

SocialFocus (https://apps.apple.com/us/app/socialfocus-hide-distractions/id1661093205) is a browser extension offered on Safari, Chrome, and Firefox. The application allows users to customize what additive elements are included on seven social media sites. Additive elements refer to the type of content displayed on the interface, and examples include comments, stories, shorts, home feeds, and premium ads. The product offers a one-time purchase but also has a pro version that can be subscribed to. This version includes hotkey configuration and settings synchronization. Because the program still uses the platforms themselves as a point of access, it is considered a helper app. They also collect certain PII to ‘provide a better service’ and do not sell such data to third parties. Lastly, the tool seems indifferently interoperable with all platforms. This is especially interesting since the app aims to minimize platform engagement, thereby reducing revenue.

Unfollow Everything 2.0

Unfollow Everything 2.0 is an experimental tool that allows Facebook users to unfollow everyone simultaneously to see if they enjoy their experience more. It is unique in the sense that it has not been released, yet it still should be included as an example as it offers unique affordances and has already been developed. It has not been released because Facebook threatened legal action against similar tools in the past, so its developer, Ethan Zuckerman, has sought declaratory relief against Meta in California court to release the product, partly to provide a space for similar tools to operate. Now, a court ruled he would have to release the product before he could seek relief.

Zuckerman claims he has the right to release the tool mainly because of a new reading of Section 230. This law has historically allowed platforms legal liability from third-party content. This new reading is based on clause (c)(2)(B), a relatively unused provision that allows an interactive content provider to restrict access to material a user might not want to be exposed to. Meta subsequently filed a motion to dismiss, contending that the plaintiff does not meet the requirements for declaratory relief because the product has not been released, and even if it did, the program would violate other laws. Zuckerman subsequently filed an opposition motion refuting these claims. This app is meant for personal use, is free, and is a helper app with server-side access. Because Unfollow Everything 2.0 and Meta are currently under adjudication, this application can be considered adversarial interoperable. It makes sense why Meta does not like this product, as it is meant to divert traffic while also having server-side access.

Tournesol

Tournesol (https://tournesol.app/) is an open-source, participatory research platform where users compare YouTube videos on Tournesol’s website to create a community-based recommendation algorithm. This algorithm is accessible while using YouTube with Tournesol’s browser extension. Because Tournesol’s goal is to “advance research in the ethics of algorithms and recommendation systems,” the product does not make revenue, although they do take donations. Interestingly, Tournesol would be considered a helper app and an alternative client. This is because a user can watch and compare videos on Tournesol’s website (alternative client) and watch the community-recommended content on YouTube with a browser extension (helper app). Because the app is open source, they will make a user’s comparisons on their site public unless specified otherwise. The product also has client-side access. I could not find any record of YouTube taking action against the software, even though it is reasonable to say that Tournesol has a negative effect on YouTube’s revenue since it changes the content consumed by depersonalizing the algorithm in favor of a community-based algorithm. A community-based algorithm can help solve the fact that “…battles over social media are rooted in the sense that the people and processes governing online spaces are unaccountable to the communities that gather in them.”

Bluesky Custom Feeds

Understanding custom feeds that interoperate with Bluesky (https://bsky.app/) requires understanding Bluesky – a federated social media platform. Federated means anyone can run a server that interoperates with the network, much like Gobo. These servers come in the form of custom feeds. Let’s take my custom feed as an example, “UMass and only”UMass.””

Every post in this feed contains the word UMass preceded by a space. I made this feed using SkyFeed, a middleware product that allows users to create feeds. I found this application under the “Community Showcase” section of Bluesky’s website, a marketplace of free third-party software containing 145 programs, many of which allow me to make an extraordinary feed where I can find all things “UMass.” In fact, I could create interoperable software myself using Bluesky’s API just by clicking the “Submit your project here” button on their website. Since Bluessky grants API access, all these custom feed applications are cooperatively interoperable with the platform and have server-side access. These products also do not make money. The data collection is a little more complicated. They run on a federated model. This means anyone can make a server, but that server’s data is connected to the platform (Bluesky). So, every custom feed product can collect your data differently. Yet, Bluesky, which accesses the servers of these custom feed products, collects and sells PII to third parties.

Bluesky is a great example of how middleware is more easily utilized in the fediverse than with VLOPs. As the Foundation for American Innovation and the Georgetown University School of Public Policy wrote, “For middleware to thrive in the present largely centralized environment, platforms must permit third-party services to operate and enable users to choose between them. However, the growing rise of federated platforms, such as Mastodon and Bluesky, and the increasing participation of major platforms in the fediverse (e.g., Meta’s Threads) creates new opportunities for the development and adoption of middleware as an integral part of the user experience.” Yet outside of Bluesky there are not a lot of middleware tools in fediverse that exist. More middleware on federated platforms would require such platforms to scale.

Characteristics of User Experience Tools

User experience tools have a wide range of characteristics, yet all of these products function by improving user experience in some capacity.

User Experience Tools Characteristics
Number of compatible apps	Varies
Use cases	Personal
Revenue model	Varies
Alternative client or helper app	Alternative client or helper app
Data collected	Varies
Client or server-side access	Varies
Type of interoperability	Adversarial, Indifferent, or cooperative
Effect on platform’s revenue?	Varies
Change in the content consumed/produced?	Consumed

Number of compatible apps. The number of compatible apps varies. For example, SocialFocus is compatible with seven platforms, while Narwhal 2 for Reddit is compatible with one.

Use cases. These programs are designed for personal use, as these applications enhance a user’s experience on a platform.

Revenue model. Most user experience tools are subscription-based, but as Unfollow Everything 2.0 demonstrated, this is unnecessary. Tournesol also only accepts donations, which is probably meant to improve the product. Products in Bluesky’s community showcase are also free.

Alternative client or helper app. Many of these programs are helper apps, but as Tournesol has shown, a user experience tool can also be an alternative client.

Data collected. Most of these applications collect PII. Unfollow Everything 2.0 does not. On the Bluesky marketplace, it varies.

Type of interoperability. If the product is a helper app, it is indifferently or adversarially interoperable. If the application is an alternative client, it tends to be cooperatively interoperable because of its API subscription.

Effect on platform’s revenue. This varies. For example, Socialfocus, Unfollow Everything 2.0, and Tournesol have a negative effect on a platform’s revenue, but Narwhal has a positive effect.

Legal implications of existing middleware

Why do platforms not integrate features middleware offers?

Middleware can ultimately be seen as a feature request. If the platform offered the features the middleware product had, there would be no market for that application. So why don’t platforms just integrate affordances middleware offers? The answer, as demonstrated in the tools sections, is that many middleware products’ functionality exists to undermine a platform’s terms of service (ToS) or divert traffic. For example, personal analytics tools allow a user to see who blocked them, in violation of Instagram’s ToS. SocialFocus is designed to divert traffic by allowing users to take away parts of the interface on a variety of platforms. Therefore, middleware products might not integrated due to harm to the user, advertisers, influencers, and, therefore, the platform itself.

One can surmise that there are more righteous reasons than others for platforms not to integrate middleware features on their product. But why a platform won’t integrate allows one to understand how functionality correlates to the legal hurdles a product might face. To make middleware a healthy tool for social media reform, an environment must be created where the good tools platforms do not want to allow should be allowed, and the bad tools should not be.

What products should platforms be forced to interoperate with?

It is reasonable to allow for good middleware products and not allow bad ones. But what applications are good? One way to think about good middleware applications, are as follows: those that provide a net benefit for users, advertisers, influencers, and non-users.

I created a sample stakeholder analysis to determine which tools follow such a definition. Obviously, there are many ways to create a stakeholder analysis that follow different definitions than this one. This way can be helpful because VLOPs have historically had relatively large amounts of agency in arbitrating control over its clients without the structural market mechanisms needed to hold VLOPs accountable to it’s clients. Therefore, basing a stakeholder analysis on these actors allows for a regulatory framework that gives power to the clients of social media platforms, not the platform itself.

Before I get to this sample stakeholder analysis, I will elaborate on the above definition. Users, advertisers, influencers, and non-users are the only actors considered because the number of people who occupy these groups and are affected by a middleware product is large enough to render other potential stakeholders, such as the middleware application itself, quantifiably negligible. For example, the number of YouTube users potentially affected by a product vastly outweighs the number of middleware employees affected by their own product.

As for the definition of these four groups, users are those who use the middleware application and/or platform on which the tool is interoperable in the US. Advertisers are those that advertise on the platform the middleware is interoperable within the US. Influencers are content creators on the platform the middleware is interoperable within the US. Non-users refer to everyone in the US, even users, as it really means the effect of a product on everyone as citizens. Net benefit is the impact on users, advertisers, influencers, and non-users. Impact means the current or future effects that are probabilistically relevant that a middleware application has.

Methods for stakeholder analysis

Operationalizing the definition of ‘good middleware applications’ can be achieved by incorporating the characteristics of middleware products from the previous section. Not every characteristic will have an effect on any stakeholder. Also, not every characteristic has the potential to harm the same actor. For example, undermining a platform’s revenue can hurt an advertiser, while accessing private data in the form of PII could help an advertiser and hurt a user. Furthermore, every actor will be harmed differently, so when operationalizing harm, it can be put on a scale of -5 to 5; -5 = most harmful, and 5 = most beneficial. I will call this the “Stakeholder Effect Score (SES).”

Users, influencers, advertisers, and non-users are also worth a different amount, proportional to the their population on the platform the middleware accesses. Calculating this entails including a “stakeholder weight SW,” which will be out of 1. The SES times the SW of an actor, plus the SES times the SW of another actor will equal the “Overall Effect Score of Characteristic (OESC).” Here is an example equation:

OESC_{Data Collected} = (SES_Userx SW_User) + (SES_Advertiserx SW_Advertiser) + (SES_Influencerx SW_Influencer) + (SES_Consumerx SW_Consumer)

The next step is to add each characteristic to create an “Overall Effect Score of Product (OESP).”

OESP_Example = OESC_{Data Collected} + OESC_{Effect on platforms revenue}

Now that an equation is in place, I will create a hierarchy of values for common stakeholders to better assess the potential harm and benefit of a given characteristic. The order of values will be based on its largest potential effect. Therefore, the highest value will have a 5/-5 SES, the next highest 4/-4 SES, and so on. For users, it can be assumed that safety is the most critical trait of a product, as a person naturally prioritizes safety over everything else. Privacy follows this: without it, there is potential for damage to one’s positionality in life when the most intimate information gets into the wrong hands. Mental health is next, as there is no direct danger or immediate consequence of one’s life prospects being ruined, yet there is still psychological distress caused. Then, it is access to information. While an increase or decrease in one’s consumption of information can change a user’s satisfaction while using the product, even if it does not, it still affects the societal discourse as a whole. Lastly is user experience (UX), which usually cannot substantially affect a user’s life. Advertisers value profit first as it is a business and control over content second. For influencers, the most critical value is safety because they are still a user, followed by profit as they are a business as well, and privacy. For non-users, I will say safety is most important, followed by privacy, economic impact, political impact, and cultural impact.

It’s essential to note that this is a rough analysis meant to serve as a model for a more vigorous examination. For example, below, I calculate the number of employees that work for a company that advertises on social media in the US to find the SW of advertisers. I could not find this directly, so I found the percentage of companies that advertise on social media in the US times the number of people employed in the US. Obviously, this is not an accurate way to find the number of employees working for a company that advertises in the US, but it is the best way available. Also, when calculating the SES, as illustrated in the above paragraph, quantifying harm is just based on logical reasoning and is therefore arbitrary; for example, safety > privacy for users.

Sample stakeholder analysis

Now, I will do a sample stakeholder analysis of Block Party. Quick disclaimer, the version of the application I am using is the one that allowed users to block harassers en masse, as this is a simpler product to start out with. The only relevant characteristic for this version of Block Party is “Change in the content consumed/produced.’ Users are also the only stakeholder because it is impossible to tell the impact for advertisers, influencers, and non-users. For instance, I do not know if safety features driving traffic up get offset by harassers using the platforms less, so I do not know how the platform affects advertisers and influencers. Therefore, the SW for users = 1. SES for users = 5 because the product can potentially save a user’s life. So, here is the equation:

OESC_{Change in the content consumed/produced} = (SES_Userx SW_User)

OESC_{Change in the content consumed/produced} = (5 x 1)

OESP_{Block Party} = OESC_{Change in the content consumed/produced}

OESP_{Block Party} = 5

Block Party is a relatively simple stakeholder analysis. Now I will apply the same method to a Grayjay, a more complicated example. Let’s start with the same characteristic, ‘Change in the content consumed/produced.’ For users, Grayjay streamlines content and gives them control over their algorithm. Because UX is seen as the least significant value, yet the UX is meaningfully altered, I will make the SES for users = 1. For advertisers, Grayjay does not produce content at all. Therefore, the SES score for advertisers = -4 because they cannot profit. The score is not -5 because 5/-5 was established for users in the Block Party example as pertaining to safety. With influencers, they no longer have their content recommended by YouTube. Still, users have an easier time finding influencers to whom they are subscribed on Grayjay. Also, influencers on smaller video-sharing platforms are exposed more on Grayjay. Therefore, I will say the SES for influencers = 2. Lastly, non-users are unaffected by a change in the content consumed/produced, so the SES for non-users = 0.

The SW requires finding the number of people in each stakeholder category divided by the total number of people occupying every stakeholder category. There are 238 million U.S. YouTube users. Finding a very rough estimate for first advertisers requires dividing $22 billion in annual ad revenue on YouTube in the US divided by the $3650 average amount U.S. businesses spend on YouTube ads which equals 6.027 million businesses. Next, multiply 6.027 million businesses by the 23.7 average number of employees per U.S. business, which equals 142.849 million employees. The number of U.S. influencers on YouTube with over five thousand followers is 385 thousand. Lastly, there are 341.262 million non-users. Adding all this makes a total of 722.496 million. Therefore, the SW for users = .329 (238 / 722.496), the SW for advertisers = .198 (142.849 / 722.496), the SW for influencers = .001 (385 / 722496), and the SW for non-users = .472 (341.262 / 722.496). Non-users are still accounted for in the SW for this characteristic even though their SES = 0 because they are harmed in the following characteristic. If they were not accounted for here when adding the OESCs to create the OESP, the OESC of this characteristic would have a disproportionately large OESC score. Now, I will apply these scores to the equation.

OESC_{Change in content consumed/produced} = (SES_Userx SW_User) + (SES_Advertiserx SW_Advertiser) + (SES_Influencerx SW_Influencer) + (SES_Consumerx SW_Consumer)

OESC_{Change in content consumed/produced} = (2 x 0.329) + (-4 x 0.198) + (2 x 0.001) + (0 x 0.472)

OESC_{Change in content consumed/produced} = (0.658) + (-0.792) + (0.002) + (0)

OESC_{Change in content consumed/produced} = -0.132

The other characteristic in the Grayjay analysis is ‘Effect on platform revenue.’ Explaining the harm score for each stakeholder in a way that makes sense requires me to go in a different order. Grayjay has a negative effect on YouTube’s revenue by taking advertisements away. Because YouTube makes less money from ads, they would have less revenue to share with influencers from said ads. This means that a negative effect on platform revenue has a negative effect on influencer revenue. If Grayjay succeeds and/or inspires copycats, this effect could be substantial. Therefore, I will make the SES for influencers = -4. When influencers have less revenue, there is less incentive for influencers to produce content. This also disincentivizes advertisers from producing content. So, the SES for advertisers = -4 as well. Because the negative effect on platform revenue causes less content to be produced, the SES score for users = -2 since access to information has the potential to be meaningfully altered and is placed as the second least important value. For non-users, three values, economic, political, and cultural impact, are potentially affected by less content production, which is caused by the negative effect on platform revenue. Because privacy and safety, the two most important values are not affected, the SES score for non-users = -3. Now, I will apply these SES scores and add the two OESCs to get the OESP.

OESC_{Effect on platforms revenue}= (SES_Userx SW_User) + (SES_Advertiserx SW_Advertiser) + (SES_Influencerx SW_Influencer) + (SES_Consumerx SW_Consumer)

OESC_{Effect on platforms revenue}= (-2 x 0.329) + (-4 x 0.198) + (-4 x 0.001) + (-3 x 0.472)

OESC_{Effect on platforms revenue}= (-0.658) + (-0.792) + (-0.004) + (-1.416)

OESC_{Effect on platforms revenue}= -2.87

OESP = OESC_{Change in content consumed/produced}+ OESC_{Effect on platforms revenue}

OESP = -0.132 + -2.87

OESP = -3.002

Even though Grayjay has a negative score, it is still clear Grayjay has some positive features, like allowing users to streamline content from multiple platforms while retaining more control over their algorithm and helping influencers on smaller platforms. Therefore, it is imperative to allow for middleware products that retain positive features and get rid of negative ones to have a positive score. That is the work of the next section.

Creating a Healthy Future for Middleware

As articulated earlier, state and federal regulation has passed and is pending which constrict the agency of users, whether through outright platform bans or bans for those under a certain age. By regulating and enabling middleware, an environment can be created that does the opposite–expand the agency of users. To evaluate the boundaries needed to create a healthy future for middleware, I will revisit some of the more complex examples from the taxonomy and offer some possible suggestions for what should not be allowed and how.

Grayjay. As attested by the last section, the main problem with Grayjay is taking away advertisements where it hurts influencers. Therefore, a middleware application that interoperates on a social platform shall not be allowed to take away ads that directly harm the revenue stream of influencers.

Most Personal Analytics Tools. There are two issues with these types of tools: showing users analytics on users who interact with them that are not provided by the platform, and collecting these other users’ PII. Addressing the first problem should require a middleware application that interoperates on a social media platform to not provide information to a client about other users on a platform that is without the consent of the platform. The second predicament is a little more tricky to solve.

Without middleware providers being able to access the information of other users, their potential use would be somewhat limited. However, allowing any product access to other users’ data also raises legitimate privacy concerns. And allowing platforms to dictate which tools get to access other users’ information allows said platform to keep gatekeeping what applications exist. There are two potential solutions to this problem. The first is using crowdsourced community standards that platforms can choose to adopt. The second involves “industry standards and formal certification processes, such as the WebTrust/CA Browser Forum or TrustArc, which can include independent auditing to ensure compliance.” Both would have their own problems, but are worth looking at.

Redact.dev. For Redact.dev, the issue is deleting content that can be used as evidence. Therefore, a middleware application that interoperates on a social platform shall not be allowed to delete a user’s information when such information can reasonably be seen as legally relevant.

Within the above suggestions, Congress could possibly create an enabling environment for middleware by forcing platforms to grant API access to middleware applications that do not break boundaries like these in the same way that federated platforms already do. This will allow middleware to have a more robust ability to meaningfully change a users experience in ways such as changing the algorithm, limiting data collection, and moderating content.

Pro-Competitive Solutions

While many mainstream reform solutions are oriented towards regulation, there are some solutions, similar to middleware, that are meant to give users more control over there experience by introducing competition. Antitrust law is one such method, and is being used in FTC v. Meta. The FTC accused the platform of engaging in anticompetitive conduct by purchasing Instagram and WhatsApp. These purchases are anti-competitive because new entrants cannot compete when Meta gains market share. This is because the purchases give Meta a monopoly over usage data collection that monetizes social advertising. Once market share is gained, new entrants cannot gain access to advertisers as online platforms are two-sided markets that also reap the benefits of network effects. These compounding problems make it virtually impossible for an entrant to gain users without changing the social mechanics offered. For these reasons, the FTC wants to force Meta to divest from Instagram and WhatsApp.

Yet it is questionable whether Meta’s divestiture from WhatsApp and Instagram will remedy problems end users have with platforms, particularly content moderation, privacy, and addiction. The FTC claims that Meta is using its two-sided market and network effects to gain market share because doing so is the most profitable way to compete due to the market structure of online platforms. Therefore, any rational online platform will try to take advantage of a two-sided market and network effects. Exploiting this market structure to one’s advantage necessitates a lack of privacy, as collecting usage data to sell to advertisers is profitable. Furthermore, it demands addictive algorithms that implement variable ratio reinforcement schedules and incentivize extreme content to retain users and gain the advantage of network effects. Therefore, if WhatsApp and Instagram were their own platforms, they could still behave in a manner that harms consumers. It is reasonable to say that antitrust law almost acts as a sledgehammer by pushing for divestiture, whereas social media needs more subtle and market-driven reforms. Also, it is debateable whether the FTC should be able to define what harms consumers beyond quantitative economic standards like price. For example, the FTC could say a platform’s monopoly harms consumer’s free speech interests, yet these interests are dependent on the person, making this metric necessarily political.

Platform interoperability is another potential solution. This solution is similar to server-side middleware interoperability (which is needed to create a more vibrant middleware ecosystem). However, platform interoperability would allow a user to access a social media application from a separate platform, not a middleware product. Proponents say platform interoperability can enable competitors to break into the currently inaccessible social media market because of the network effects problem. Platform interoperability is already used in the “ActivityPub,” a protocol that allows separate servers with complete autonomy to interoperate with each other. Mandating interoperability between social media platforms is promising. Still, there are questions such as which features would be interoperable and how data would be shared between companies with diverse data practices due to differences in data encryption, retention, and state laws. The differences in data practices would also need to be communicated to users seamlessly, which is logistically challenging.

Finally, there is data portability, which would allow a user to transfer data from a social media platform to its competitor. It is being suggested as a way for new entrants to access the social media market. This is because if a platform already knows one’s preferences because they collect their data, a user would be disincentivized to switch to another application without data portability. After all, the new platform is not as advanced algorithmically. Facebook has already implemented this concept to comply with EU and California law, with Facebook allowing users to download their data. However, this has not facilitated competition, and the data is in Facebook’s format, which can change anytime. Furthermore, like with middleware, there are questions of consent for importing downloaded data, which inevitably downloads a customer’s interactions with other users.

Conclusion

There is a wealth of products offering a diverse range of affordances in the social media middleware space, from those that change a users feed to applications used by companies as big as Microsoft. Categorizing these varied programs helps us understand their unique patterns, purposes, and functionality. This is needed for middleware to be considered as an alternative avenue for social media reform. Because although I defined middleware as third-party products that access a social media platform’s data to offer users affordances that differ from a platform’s default affordances, in reality, these ‘affordances’ include everything from changing an interface to blocking a harasser.

Because middleware does so much and has the potential to do much more, it is essential to make it an avenue for reform. While other pro-competitive to social media reform can help, like middleware, they are not silver bullets. And if middleware is to be utilized, its applications should do more good than harm. That is why a stakeholder analysis is important: it helps gauge what middleware products and features are helpful and harmful. The stakeholder analysis also helps form a regulatory framework that can consider all the major actors involved.

Such a framework is needed, as laws or litigation seem like prerequisites to creating a healthier social media space through middleware. This is because, as seen in the taxonomy, existing middleware programs can be harmful, and helpful middleware products, which often require API access, are not necessarily allowed by the platform.

All this is to say that creating a healthy and robust social media middleware landscape will be difficult, as helpful programs occupy the same internet as harmful ones. Still, many products have shown potential to develop this space, which should be enough to utilize middleware as one tool to help solve the social media problem.

Works Cited

ACLU of Florida Media Office. “ACLU of Florida Statement on Gov. DeSantis Signing Social Media Censorship Bill into Law.” ACLU of Florida, 26 March 2024, https://www.aclufl.org/en/press-releases/aclu-florida-statement-gov-desantis-signing-social-media-censorship-bill-law. Accessed 5 March 2025.

Allyn, Bobby. “U.S. bans TikTok unless it is sold.” NPR, 24 April 2024, https://www.npr.org/2024/04/24/1246663779/biden-ban-tiktok-us. Accessed 20 August 2024.

Apple. “Helping – Protect Kids.” Apple Developer, 2025, https://developer.apple.com/support/downloads/Helping-Protect-Kids-Online-2025.pdf. Accessed 5 March 2025.

Ark, Tom Vander. “Social Media Jacked Your Attention And Sold It, How To Get It Back And Thrive.” Forbes, 13 May 2019, https://www.forbes.com/sites/tomvanderark/2019/05/13/social-media-jacked-your-attention-and-sold-it-how-to-get-it-back-and-thrive/. Accessed 23 August 2024.

ASIAN TURBO LIMITED. “Panda Spy – For your Networks on the App Store.” App Store, 2024, https://apps.apple.com/us/app/social-spy-for-your-networks/id1107522062. Accessed 20 August 2024.

Block Party. “Does Block Party collect my data?” Block Party, 2024, https://www.blockpartyapp.com/product-faq-articles/does-privacy-party-collect-my-data/. Accessed 16 October 2024.

Bluesky. “Community Showcase.” Bluesky, 2025, https://docs.bsky.app/showcase?tags=client. Accessed 6 February 2025.

Bose, Sweta. “The Risks of Using Public Wi-Fi to Access Instagram.” Cyber Security News, 10 October 2023, https://cybersecuritynews.com/the-risks-of-using-public-wi-fi-to-access-instagram/. Accessed 13 August 2024.

Britannica. “Arab Spring | History, Revolution, Causes, Effects, & Facts.” Britannica, 20 August 2024, https://www.britannica.com/event/Arab-Spring. Accessed 2 September 2024.

Buckbee, Michael. “What is PGP Encryption and How Does It Work?” Varonis, 2 June 2023, https://www.varonis.com/blog/pgp-encryption. Accessed 20 September 2024.

Buffer. “A transparent company since 2010.” Buffer, 2024, https://buffer.com/open. Accessed 24 August 2024.

Buffer. “Why your data in Buffer might differ from your native analytics.” Buffer Help, 30 July 2024, https://support.buffer.com/article/525-why-your-data-in-buffer-might-differ-from-your-native-analytics. Accessed 21 August 2024.

Ceci, Laura. “U.S. YouTube annual revenue 2025.” Statista, 9 December 2024, https://www.statista.com/statistics/1434277/annual-revenue-of-youtube-usa/. Accessed 29 January 2025.

Census.gov. “U.S. and World Population Clock.” U.S. Census Bureau, 2025, https://www.census.gov/popclock/. Accessed 28 January 2025.

Chan, Kelvin. “Meta fined record $1.3 billion and ordered to stop sending European user data to US.” AP News, 22 May 2023, https://apnews.com/article/meta-facebook-data-privacy-fine-europe-9aa912200226c3d53aa293dca8968f84. Accessed 20 August 2024.

Chilson, Neil, and Casey Mattox. “[The] Breakup Speech: Can Antitrust Fix the Relationship Between Platforms and Free Speech Values?” | Knight First Amendment Institute, 5 March 2020, https://knightcolumbia.org/content/the-breakup-speech-can-antitrust-fix-the-relationship-between-platforms-and-free-speech-values. Accessed 2 September 2024.

Cloudfare. “What is rate limiting? | Rate limiting and bots.” Cloudflare, https://www.cloudflare.com/learning/bots/what-is-rate-limiting/#. Accessed 4 September 2024.

Confessore, Nicholas. “Cambridge Analytica and Facebook: The Scandal and the Fallout So Far (Published 2018).” The New York Times, 4 April 2018, https://www.nytimes.com/2018/04/04/us/politics/cambridge-analytica-scandal-fallout.html. Accessed 20 August 2024.

Cornell Legal Information Institute. “47 U.S. Code § 230 – Protection for private blocking and screening of offensive material.” Law.Cornell.Edu, 2018, https://www.law.cornell.edu/uscode/text/47/230. Accessed 21 August 2024.

Cyd. “Claw back your data from Big Tech.” cyd.social, 2024, https://cyd.social/. Accessed 6 March 2025.

Doctorow, Cory. “Adversarial Interoperability: Reviving an Elegant Weapon From a More Civilized Age to Slay Today’s Monopolies.” Electronic Frontier Foundation, 7 June 2019, https://www.eff.org/deeplinks/2019/06/adversarial-interoperability-reviving-elegant-weapon-more-civilized-age-slay. Accessed 5 October 2024.

Fonsi, Louis. “YouTube Statistics 2024 [Users by Country + Demographics].” Global Media Insight, 3 October 2024, https://www.globalmediainsight.com/blog/youtube-users-statistics/. Accessed 21 October 2024.

Fonsi, Louis. “YouTube Statistics 2025 [Users by Country + Demographics].” Global Media Insight, 9 January 2025, https://www.globalmediainsight.com/blog/youtube-users-statistics/. Accessed 29 January 2025.

Freedom Apps. “Privacy Policy.” Freedom Apps, 2024, https://freedom-apps.com/privacy. Accessed 21 January 2024.

Friendly. “Friendly Privacy Policy.” Sensor Tower, 5 January 2024, https://sensortower.com/friendly-privacy-policy. Accessed 26 August 2024.

Friendly Social. “friendly.io.” Friendly, Simple Apps for iOS and Android, 2023, https://friendly.io. Accessed 13 August 2024.

Fukuyama, Francis, et al. “MIDDLEWARE FOR DOMINANT DIGITAL PLATFORMS: A TECHNOLOGICAL SOLUTION TO A THREAT TO DEMOCRACY.” Stanford Cyber Policy Center, https://fsi-live.s3.us-west-1.amazonaws.com/s3fs-public/cpc-middleware_ff_v2.pdf. Accessed 21 August 2024.

GopGop Apps. “Reports+ Unfollowers Followers on the App Store.” App Store, 2024, https://apps.apple.com/us/app/reports-unfollowers-followers/id1481469888. Accessed 20 August 2024.

Grayjay. “Grayjay App – Follow Creators Not Platforms.” Grayjay App – Follow Creators Not Platforms, 2023, https://grayjay.app/index.html. Accessed 20 August 2024.

Grayjay. “Grayjay Privacy Policy.” Grayjay, 18 August 2023, https://grayjay.app/privacy-policy.html. Accessed 23 August 2024.

Haidt, Johnathan. The Anxious Generation: How the Great Rewiring of Childhood Is Causing an Epidemic of Mental Illness. Penguin Publishing Group, 2024. Accessed 5 March 2025.

Hogg, Luke, and Renée DiResta. “Shaping the Future of Social Media with Middleware.” Foundation for American Innovation, 17 December 2024, https://www.thefai.org/posts/shaping-the-future-of-social-media-with-middleware. Accessed 16 February 2025.

Hootsuite. “Benefits of Hootsuite: Social Media Management Tools and Software.” Hootsuite, 2024, https://www.hootsuite.com/why-hootsuite. Accessed 9 August 2024.

Idpiumass. “The Three-Legged Stool: A Manifesto for a Smaller, Denser Internet.” Initiative for Digital Public Infrastructure at UMass Amherst, 29 March 2023, https://publicinfrastructure.org/2023/03/29/the-three-legged-stool/. Accessed 25 September 2024.

Influencity. “The Largest Influencer Study in the US in 2023.” Influencity, https://influencity.com/resources/studies/the-largest-influencer-study-of-the-united-states-2023/. Accessed 28 January 2025.

Kang, Cecilia. “Surgeon General Calls for Warning Labels on Social Media Platforms.” The New York Times, 18 June 2024, https://www.nytimes.com/2024/06/17/health/surgeon-general-social-media-warning-label.html. Accessed 5 March 2025.

Knight First Amendment Institute. “Zuckerman v. Meta Platforms, Inc.” Knight First Amendment Institute, 2024, https://knightcolumbia.org/cases/zuckerman-v-meta-platforms-inc. Accessed 14 September 2024.

Lane, Spencer. “Gobo 2.0: All Your Social Media in One Place.” Initiative for Digital Public Infrastructure at UMass Amherst, 9 November 2022, https://publicinfrastructure.org/2022/11/09/gobo-2-0-all-your-social-media-in-one-place/. Accessed 7 August 2024.

Laura Ceci. “U.S. YouTubers by followers 2023.” Statista, June 2023, https://www.statista.com/statistics/1398231/us-youtubers-by-followers/#:~:text=As%20of%20June%202023%2C%20almost,as%20of%20the%20examined%20period. Accessed 29 January 2025.

Liu, Henry. “Facebook, Inc., FTC v.” Federal Trade Commission, 12 September 2024, https://www.ftc.gov/legal-library/browse/cases-proceedings/191-0134-facebook-inc-ftc-v. Accessed 1 October 2024.

Livintis W.L.L. “Repost – Privacy Policy — Livintis.” Livintis, December 2019, http://www.livintis.com/repost-privacy-policy. Accessed 29 January 2025.

Meta, “Insights – Instagram Platform – Documentation – Meta for Developers.” Meta, 2025, https://developers.facebook.com/docs/instagram-platform/insights/. Accessed 5 March 2025.

Messenger for WhatsApp Duo Web. “Privacy Policy for Messenger for WhatsApp Duo Web.” freeprivacypolicy.com, 3 May 2024, https://www.freeprivacypolicy.com/live/d6c4e431-9e4c-4494-88cd-1e7e178f078c. Accessed 23 August 2024.

Orben, Amy, and Andrew K. Przybylski. “The association between adolescent well-being and digital technology use.” Nature, 14 January 2019, https://www.nature.com/articles/s41562-018-0506-1. Accessed 15 April 2025.

Osoba, Stella, and Peter Gratton. “12 States With Teens’ Social Media Regulation- Is Yours One of Them?” Investopedia, 11 December 2024, https://www.investopedia.com/states-with-social-media-regulation-for-teens-8757983. Accessed 5 March 2025.

Practising Law Institute. “Social Media and Mobile Devices 2024: Addressing Corporate Risks NY.” PLI, 2024, https://www.pli.edu/programs/social-media?t=ondemand. Accessed 28 January 2025.

Rajendra, Chand, and Ethan Zuckerman. “An Illustrated Field Guide to Social Media | Knight First Amendment Institute.” | Knight First Amendment Institute, 14 May 2021, https://knightcolumbia.org/blog/an-illustrated-field-guide-to-social-media. Accessed 26 August 2024.

Reckeweg, Anika. “Axel Springer vs. Adblock Plus: Federal Court of Justice awaits ECJ ruling.” Heise, 26 July 2024, https://www.heise.de/en/news/Axel-Springer-vs-Adblock-Plus-Federal-Court-of-Justice-awaits-ECJ-ruling-9813824.html. Accessed 6 October 2024.

Redact.dev. “Redact Privacy App – Overview of Supported Services for Mass Deletion.” Redact, 2024, https://redact.dev/services. Accessed 20 August 2024.

Restream. “Kick vs. Twitch: Full Comparison.” Restream, 2024, https://restream.io/learn/comparisons/kick-vs-twitch/. Accessed 21 October 2024.

Rossmann, Louis. “Youtube’s Legal Team sent me a letter! 😃.” YouTube, 1 May 2024, https://www.youtube.com/watch?v=dqTYg6vnQvw. Accessed 20 August 2024.

RSS.app. “Privacy Policy.” RSS.app, 24 November 2019, https://rss.app/privacy. Accessed 4 September 2024.

RSS.app. “RSS Feed Generator, Create RSS feeds from URL.” RSS.app, 2024, https://rss.app. Accessed 4 September 2024.

SocialFocus. “SocialFocus | Privacy Policy.” SocialFocus, 2024, https://socialfocus.app/privacy/. Accessed 6 September 2024.

Spicy Apps. “Privacy Policy.” Spicy Apps, 25 May 2018, https://www.spicyapps.com/privacypolicy.htm. Accessed 30 January 2025.

Spicy Apps. “Unfollow for X aka Twitter on the App Store.” App Store, 2024, https://apps.apple.com/us/app/unfollow-for-twitter/id369327398. Accessed 20 August 2024.

Sprinklr. “Personally Identifiable Information & Privacy Compliance.” Sprinklr, 2024, https://www.sprinklr.com/help/articles/pii-masking/personally-identifiable-information-privacy-compliance/6463297330f12540268fa9b8#. Accessed 29 January 2025.

Statista. “Rumble global MAUs 2024.” Statista, 13 August 2024, https://www.statista.com/statistics/1347599/rumble-quarterly-mau/. Accessed 21 October 2024.

Statista Research Department. “Average number of employees per primary firms U.S. 2027.” Statista, 8 January 2025, https://www.statista.com/statistics/1411501/average-number-of-employees-per-primary-firms-us/. Accessed 29 January 2025.

Stokel, Chris. “Under Elon Musk, X is denying API access to academics who study misinformation.” Fast Company, 27 February 2024, https://www.fastcompany.com/91040397/under-elon-musk-x-is-denying-api-access-to-academics-who-study-misinformation. Accessed 23 August 2024.

Stoltz, Mitch, and Andrew Crocker. “Once Again, Facebook Is Using Privacy As A Sword To Kill Independent Innovation.” Electronic Frontier Foundation, 20 November 2020, https://www.eff.org/deeplinks/2020/11/once-again-facebook-using-privacy-sword-kill-independent-innovation. Accessed 5 September 2024.

Stray, Johnathan, et al. “The Algorithmic Management of Polarization and Violence on Social Media.” Knight First Amendment Institute, 22 August 2023, https://knightcolumbia.org/content/the-algorithmic-management-of-polarization-and-violence-on-social-media. Accessed 16 February 2025.

Tournesol. “Tournesol.” Tournesol, 2021, https://tournesol.app. Accessed 5 February 2025.

Tretiakov, Yevhen. “SocialFocus.” SocialFocus, https://socialfocus.app. Accessed 6 September 2024.

U.S. Census Bureau. “U.S. and World Population Clock.” U.S. Census Bureau, 2025, https://www.census.gov/popclock/. Accessed 29 January 2025.

WebFX. “How Much Does YouTube Advertising Cost in 2025?” WebFX, 13 December 2024, https://www.webfx.com/social-media/pricing/how-much-does-youtube-advertising-cost/. Accessed 29 January 2025.

Wharton Online. “What Is the Network Effect? – Wharton Online.” Wharton Online, 17 January 2023, https://online.wharton.upenn.edu/blog/what-is-the-network-effect/. Accessed 20 August 2024.

X Developer Platform. “Getting Started with the Twitter API | Docs | Twitter Developer Platform.” Twitter Developer Platform, 2024, https://developer.x.com/en/docs/x-api/getting-started/about-x-api#item1. Accessed 22 August 2024.

YCharts. “US Employed Persons Monthly Insights: Employment Situation.” YCharts, 10 January 2025, https://ycharts.com/indicators/us_employed_persons. Accessed 28 January 2025.

Zuckerman, Ethan. “Opinion | I’m Suing Meta So Users of Facebook Can Control the Content They See.” The New York Times, 5 May 2024, https://www.nytimes.com/2024/05/05/opinion/facebook-court-internet-meta.html. Accessed 21 August 2024.Zuckerman, Ethan, and Chand Rajendra-Nicolucci. “Let the community work it out: A throwback to early internet days could fix social media’s crisis of legitimacy.” NiemanLab, 25 October 2023, https://www.niemanlab.org/2023/10/let-the-community-work-it-out-a-throwback-to-early-internet-days-could-fix-social-medias-crisis-of-legitimacy/. Accessed 16 February 2025.