1
When members of the Cockroach Janata Party gathered at Delhi’s Jantar Mantar recently to protest, it appeared to be one of the many demonstrations routinely held in the national capital. Yet within days, videos carrying the campaign’s message began circulating across Instagram and Facebook, political commentary emerged in multiple languages, and advertisements linked to the campaign appeared in Meta’s Ad Library. What began as a small protest had acquired a far larger digital presence, raising a broader question about the role Meta’s platforms now play in shaping political discourse. An examination by Read of Meta’s public advertising records, the company’s published material on its recommendation systems and years of scrutiny by the United States Congress suggests that Facebook and Instagram have evolved well beyond conventional social networking platforms. Together, they now form one of the world’s most powerful systems for distributing political content, combining paid advertising with algorithmic recommendations capable of expanding the reach of political messaging far beyond its original audience. Meta’s own Ad Library reports that since August 2019, more than five million advertisements relating to social issues, elections and politics have been published in India, with cumulative disclosed spending exceeding Rs 778 crore or US$90.5 million.
That works out to an average of roughly Rs 30.4 lakh spent every day over the past seven years. The figures underscore the extent to which Meta’s platforms have become a major infrastructure for political advertising and public issue campaigns. During the same period, Ukraine recorded 564,800 such advertisements with spending of US$14.83 million, Bangladesh recorded 288,770 advertisements with spending of US$2.61 million, Pakistan recorded 137,197 advertisements with spending of approximately US$1.6 million, while Nepal recorded 39,959 advertisements with disclosed spending of US$367,613. Political mobilisation has historically followed a familiar pattern. An issue emerges, organisations mobilise supporters, demonstrations are held, journalists report on them and public opinion gradually develops. Digital platforms have altered that sequence. Increasingly, political narratives gain momentum online before they achieve comparable visibility on the ground. Unlike newspapers or television broadcasters, Meta’s platforms do not merely publish content in chronological order. Every time a user opens Facebook or Instagram, thousands of posts compete for limited space in that individual’s feed. The company uses machine-learning systems to predict which posts a user is most likely to find relevant and engage with.
Those predictions are based on a wide range of signals, including previous viewing behaviour, comments, shares, viewing time, relationships between users, interests and numerous other behavioural indicators accumulated over time. The significance of this system lies in the fact that every user experiences a different information environment. Two individuals following similar political accounts may still receive markedly different recommendations because the platform continuously personalises content according to predicted user behaviour. Meta describes this process as being driven by recommendation algorithms, which use machine-learning systems to rank and recommend content based on predicted user relevance and engagement. Meta says that the content shown on Facebook and Instagram is primarily ranked and recommended by machine-learning systems rather than by employees manually deciding what each individual user sees. Every time a user opens the app, the system evaluates thousands of possible posts using signals such as previous activity, viewing time, shares, comments, relationships and predicted relevance before determining what is most likely to appear in that person’s feed. In that sense, the ranking of content is automated. However, the algorithms do not operate independently of human decisionmaking. Meta’s engineers design and update the recommendation systems, determine the objectives they optimise for, and establish the policies that govern what content is eligible for recommendation.
The company therefore does not manually select posts for individual users, but people define the rules and parameters within which the automated systems operate. Consider two political protests held in Delhi on the same day, each attended by about 500 people. Videos from one protest may receive wider distribution because the recommendation system gives greater weight to signals such as watch time and shares. If Meta later changes how its recommendation system ranks content, the relative visibility of the two protests could also change. The ranking would still be carried out automatically by the algorithm, but the rules governing that algorithm would have been determined by Meta. At the same time, because Meta develops and maintains those recommendation systems, it has the technical ability to modify the models and ranking signals over time. The company has publicly acknowledged making changes to its ranking systems through updates to its recommendation models and policies, although it does not publicly disclose the full details of every change. For example, it could decide to give greater importance to watch time than shares, prioritise content from friends over public pages, or reduce the recommendation of political content more broadly. Such changes can influence how widely different categories of content are distributed.
The existence of that capability, however, does not by itself establish that Meta has used it to favour any particular political party, protest or ideology. Congressional investigations and court records have shown that Meta has maintained regular engagement with several U.S. government agencies, including the FBI, DHS, CISA, the Office of the Director of National Intelligence and the State Department, on issues ranging from election security and foreign influence operations to cyber threats. Those interactions have become the subject of intense political scrutiny, with some lawmakers alleging that the relationship between federal agencies and social media platforms blurred the line between legitimate security cooperation and government influence over online speech. Meta has denied that government agencies dictate its content moderation decisions.
One of the few public windows into Meta’s political advertising ecosystem is the Meta Ad Library. Unlike commercial advertisements, advertisements relating to politics, elections and public issues disclose estimated expenditure, impression ranges, advertiser details and publishing entities. Although the figures are presented as ranges rather than precise numbers, they provide an unusually transparent view of political advertising on the platform. A search of the Meta Ad Library using the keyword “cockroach” under the “Issues, elections or politics” category in India identified multiple advertisements. One advertisement disclosed an estimated expenditure of between Rs 3,000 and Rs 3,500 while recording an estimated 400,000 to 450,000 impressions. In effect, an estimated expenditure of around Rs 3,000 generated between 400,000 and 450,000 impressions on Meta’s platforms. While impressions do not necessarily represent unique users, the figures illustrate how relatively modest spending can generate substantial visibility online.
Other advertisements reflected lower expenditure but still generated impression ranges extending into tens or hundreds of thousands. Although the available data does not reveal precisely how many unique individuals viewed the advertisements or subsequently engaged with them, it demonstrates that significant political visibility can be purchased on Meta’s platforms at comparatively modest cost. Paid political advertising, however, represents only one component of Meta’s distribution ecosystem. Once an advertisement reaches its intended audience, subsequent visibility depends largely on how users respond. If viewers spend time watching the content, share it, comment on it or otherwise engage with it, Meta’s recommendation systems identify those signals and recommend similar content to additional users.
The company’s published material explains that its ranking systems rely on multiple predictive signals, including anticipated engagement, to determine what appears in individual feeds. This creates a feedback loop. Paid promotion introduces political content to an initial audience. Audience engagement generates additional behavioural signals, which Meta’s recommendation systems use to show the content to more users. If engagement continues, visibility expands further through recommendations. This process should not be interpreted as evidence that Meta intentionally promotes one political ideology over another.
Nor does it establish that every political campaign experiences such amplification. Rather, it illustrates how paid distribution and engagement-based recommendations work together to increase the visibility of political content beyond its initial advertising reach. The implications for political mobilisation are significant. Historically, the perceived strength of a political movement was closely associated with the size of the crowd it could assemble. Today, a demonstration attended by hundreds of people can reach hundreds of thousands of users online within a short period. Digital visibility and physical mobilisation have become related but distinct measures of political influence.
Consequently, the prominence of a political issue on Facebook or Instagram should not automatically be interpreted as evidence of an equally large movement on the ground. Online visibility may reflect a combination of genuine grassroots support, paid promotion, user engagement, recommendation systems and social sharing. As a result, the scale of online discussion does not necessarily reflect the scale of physical mobilisation. These concerns are not new. Since the Cambridge Analytica controversy in 2018, Meta Chief Executive Officer Mark Zuckerberg has repeatedly appeared before committees of the United States Congress. Those hearings have examined election interference, political advertising, misinformation, privacy, content moderation, antitrust issues and platform safety. Although the subjects have differed, lawmakers have consistently questioned whether Meta’s unprecedented scale, sophisticated advertising infrastructure and recommendation systems possess the ability to influence democratic discourse in ways that traditional media never could.
The evidence examined by Read does not establish that Meta creates political movements, nor does it demonstrate that the company intentionally favours any political party or ideology. Political grievances continue to originate with citizens, activists and political organisations. What has changed is the infrastructure through which those grievances can be amplified. Unlike traditional political campaigns that depended largely on physical mobilisation and conventional media coverage, digital campaigns can now purchase initial visibility and, if they generate sufficient engagement, potentially reach much larger audiences through platform recommendations. Organisations no longer require access to a television network or the front page of a newspaper to reach large audiences. They require compelling digital content, an advertising budget and access to platforms whose recommendation systems can significantly extend the reach of messages that generate sustained engagement.