In recent days, a growing wave of social media users across India have expressed deep frustration as videos and posts detailing the Cockroach Janta Party’s (CJP) protests are repeatedly pushed into their feeds. This intrusive material is reportedly appearing on screens regardless of whether users follow official CJP accounts, consume political content, or have recently registered new profiles.

Detailing her experience on X, user Mohini observed, “Over the past few days, we have seen systematic manipulation of algorithms to push the CJP narrative on Meta’s social media platforms. On Instagram in particular, there appears to be a funded campaign to spread anarchy and unrest among the younger generation.”

Echoing similar concerns, another commentator, Always4Right, pointed out, “Instagram has been hijacked. It is mostly pushing an anti-police and anti-government narrative, which is creating a sense of violence against the police.”

The visible proliferation of this content prompted user Ashwini to state, “Instagram is aggressively pushing pro-CJP protest content,” while urging government authorities to step in immediately.

Further emphasizing the potential severity of the trend, X user Shubh added, “Instagram has launched full algorithm-supported propaganda for the CJP protests. The plan to turn India into Nepal or Bangladesh has been activated. It needs to be seriously countered.”

Providing deeper technical context to this phenomenon, detailed posts by commentator RevolutionMonk revealed that even users who completely avoid interacting with CJP-linked handles continue to find these short-form reels dominating their feeds. Explaining the mechanism behind this, he noted, “That is because CJP has a network of state-wise social media coordinators who actively collaborate with Instagram influencers.” He further highlighted that coordinated posting campaigns were being financially incentivized, with payments starting from Rs 5,000. In follow-up posts, RevolutionMonk published screenshots displaying identical content published across multiple Instagram accounts to substantiate his finding.

This persistent push across platforms like Facebook, Instagram, and X raises a fundamental query: Is this material going viral due to genuine user demand, or are algorithmic systems actively driving it far past its organic reach?

The Operational Architecture of Modern Social Recommendation Systems

The answer lies directly within the design of contemporary social media networks. Platforms like Facebook and Instagram no longer function purely as reverse-chronological feeds showing updates from accounts a user explicitly chooses to follow. Instead, Meta determines which political content qualifies for wide recommendation, selects which creators appear before non-followers, and evaluates which implicit behavioral signals are sufficient to conclude that a person might want to see a political campaign.

These internal parameters have been altered repeatedly. Meta itself confirmed that it has renewed efforts to actively surface political material by tracking both explicit and implicit user actions. In algorithmic terms, an explicit signal corresponds to a deliberate like or comment, whereas an implicit signal can be as simple as momentarily pausing to view a post.

Consequently, a user does not need to align with the CJP, follow its accounts, or manually look up its demonstrations. Merely pausing to watch a single dramatic clip provides enough data for Meta's underlying engine to begin populating that user’s feed with similar protest footage.

Meta’s Policy Shift on Political Content Visibility

In 2021, Meta took measures to dial back the volume of political and civic content shown to users, later reducing the weight given to comments and shares when ranking political material. At the time, the company maintained that users generally preferred feeds uncluttered by political disputes.

However, a major strategic pivot occurred in January 2025. Meta announced that it was re-introducing political content across Facebook, Instagram, and Threads using what it described as a "more personalised approach." Under this updated framework, political material is evaluated using explicit signals alongside implicit indicators like passive view time. Meta confirmed that these signals would directly drive broader recommendations of political posts.

This 2025 policy adjustment offers key context for why CJP protest reels are reaching entirely unrelated feeds.

A user might pause on a video out of alarm over an attack on police personnel, shock at the behavior of demonstrators, or simply a desire to identify where an event occurred. Meta’s automated systems do not evaluate the viewer's underlying intent; they simply record that the media successfully retained user attention.

Once a user pauses, watches a few seconds, opens the comment section, or visits the creator's profile, multiple engagement signals are registered. The system responds by recommending another protest video, followed by additional accounts covering the same political mobilization.

According to Meta’s official Help Centre documentation, suggested posts on Instagram are chosen based on a user’s prior activity, their connected network, post-specific data, and the overall popularity of the sharing account. The system weighs how other users across the platform interact with a given piece of content, surfacing suggestions within the main feed, the Explore page, and the Reels tab. Thus, high-engagement CJP footage does not wait to be sought out—it is delivered directly to users.

Why Recommendation Systems Misinterpret Negative Engagement

Social media firms routinely position recommendation engines as neutral tools responding to user preferences. However, the operational definition of "interest" applied by these systems is exceptionally broad.

A user who posts an angry comment opposing a CJP clip is still generating comment volume. A viewer who forwards a video to condemn the actions of protesters is still increasing its share count. An individual rewatching a violent clash to make sense of the incident is actively boosting its watch-time statistics. To an automated recommendation pipeline, every single one of these actions signals that the content is compelling and should be presented to a wider audience.

Empirical research highlights the structural power of these systems. Academic studies conducted around the 2020 US presidential election revealed that replacing Facebook's and Instagram's algorithmic feeds with simple chronological feeds led to a substantial reduction in total time spent on the platforms, while significantly altering the nature of political material users encountered. While researchers did not observe immediate short-term shifts in political opinions, the data proved that algorithms fundamentally restructure what users see and how they behave online.

Similarly, a comprehensive study analyzing aggregated data from over 208 million Facebook users demonstrated that political and ideological segregation increased progressively as users moved from content they could potentially see, to what the algorithm selected for them, and finally to what they engaged with.

These findings show that recommendation engines do not merely reflect the public sphere; they re-shape it by selectively distributing visibility. While CJP campaign content originates with its core organizers, it reaches much broader audiences once high-friction footage generates enough early metrics to be selected by recommendation loops.

Algorithms reflect Corporate Policy Choices

Describing "the algorithm" as an autonomous, neutral entity obscures the level of control social media platforms maintain over their software.

Meta explicitly specifies which engagement signals carry weight. It decides whether comments, shares, watch duration, negative feedback, or user survey scores should dictate political distribution. Furthermore, the company determines which categories of media are eligible to reach audiences outside an account's existing follower base.

In 2022, Meta globally decreased the weight assigned to comments and shares on political posts. By 2023, the platform stated it was moving away from pure engagement metrics in favor of user survey responses detailing what content people found informative.

By 2024, Meta restricted proactive recommendations of political posts from unfollowed accounts on Instagram and Threads, only to reverse course in 2025 by making political recommendations more personalized. Currently, Facebook's political content controls are turned on by default, requiring users to manually adjust their settings if they wish to view fewer political posts.

These continuous adjustments demonstrate that the visibility of political content is governed by deliberate corporate policy choices. Meta does not need to issue a directive to promote a specific political entity like the CJP; it simply needs to classify its high-engagement footage as recommendation-eligible political material and allow automated distribution mechanisms to take over.

Asymmetric Moderation Dynamics in Clashes Involving Law Enforcement

A key layer of this distribution controversy involves how different forms of protest footage are moderated. Identical acts of physical violence can receive starkly different treatment based on who is depicted and how the clip is captioned.

Footage showing police personnel using force against demonstrators is often framed as evidence of police heavy-handedness, civil rights violations, or state overreach. Consequently, such uploads are frequently protected and distributed under broad journalistic exceptions for newsworthy documentation.

Conversely, footage documenting protesters physically attacking law enforcement officers is far more likely to be flagged and suppressed for displaying graphic violence. If an banned or extremist group appears in the video, the post falls under strict rules regarding the representation or glorification of dangerous organizations.

While Meta’s policies on Dangerous Organisations and Individuals technically permit neutral news reporting and condemnation of violence, the boundary between journalistic coverage and glorification is frequently misapplied by automated filters and human moderators alike.

A clear example of this operational failure occurred when Meta’s automated systems removed an Urdu-language newspaper report covering the Taliban. Two human reviewers upheld the removal, resulting in account penalties for the administrator. Although the case was routed to an internal queue designed to catch improper removals, it sat unreviewed because Meta had fewer than 50 Urdu-language reviewers assigned to that queue, and the item was not categorized as high priority. The content was restored only after Meta’s independent Oversight Board selected the case for review, prompting the company to acknowledge that the post was a legitimate news report.

This case demonstrates the frequent misalignment between policy intentions and real-world moderation. While rules may officially allow for news coverage of political violence, automated systems and human reviewers often end up penalizing legitimate reporting instead of unlawful behavior.

Meta's Admissions Regarding Over-Moderation and Speech Restrictions

Meta has publicly acknowledged that its content moderation infrastructure frequently overreached. In January 2025, the company stated that under intense public and political pressure, it had built increasingly complex enforcement mechanisms that ultimately went too far. Meta conceded that substantial amounts of benign content were mistakenly censored and that numerous accounts were subjected to incorrect penalties.

To illustrate the scale, Meta revealed that during December 2024 alone, it removed millions of posts daily, estimating that between 10% and 20% of its enforcement actions may have been erroneous.

Given the billions of posts processed, even a 10% error rate represents millions of legitimate posts facing wrongful suppression or reduced reach.

Additionally, Meta acknowledged that its third-party fact-checking program frequently applied intrusive labels to legitimate political discourse, unintentionally suppressing its distribution. The company stated that the program had too often functioned as a tool for unwarranted censorship when individual fact-checkers brought personal biases to the review process.

These admissions confirm that concerns regarding systemic platform bias are rooted in acknowledged operational realities, as Meta’s current position centers on actively attempting to reduce these documented enforcement errors.

The Role of Human Moderation Teams in Shaping Online Narratives

Automated algorithms do not operate in isolation; human review teams play a direct role in content distribution. In 2024, Meta reported employing approximately 15,000 content reviewers across Facebook, Instagram, and Threads covering over 80 languages. Higher-level political moderation decisions involve policy teams, engineering groups, legal advisors, and specialized election operations centers.

Human personnel evaluate whether a clip condemns violence or promotes it, assessing whether a post serves an activist, journalistic, or satirical purpose. These reviewers are also tasked with interpreting captions written in regional languages, which may contain local political references unfamiliar to personnel operating outside those regions.

Subjectivity can influence outcomes at multiple stages: during the drafting of platform guidelines, when organizations are added to watchlists, when reviewers assess the tone of a commentary, or when coordinated reporting campaigns target specific ideological viewpoints.

As demonstrated in the Taliban reporting case, human reviewers can repeatedly confirm inaccurate moderation decisions, while regional language reviewer shortages can leave incorrect actions uncorrected for extended periods.

Documented Impact of Moderation Policies on Independent Media Reach

For digital news outlets like OpIndia, enforcement actions have measurable consequences on public reach.

In 2020, OpIndia documented a significant decline in its Facebook page reach, which dropped from approximately 20 lakh daily views to roughly 1.71 lakh. Facebook cited general content-quality concerns, explaining that automated systems had flagged rule violations. However, identical reporting content removed from OpIndia’s page remained accessible on other third-party websites without penalty.

In September 2024, Facebook restricted users from sharing OpIndia’s 187-page dossier regarding Wikipedia, labeling the link as spam and citing misleading engagement practices. The dossier itself was a standard document download that required no deceptive clicks, pop-ups, or page likes.

Subsequently, in March 2026, Meta removed an OpIndia video report that critically examined a BBC narrative regarding a Kashmiri journalist seeking housing in Delhi. The video upload was restricted within seconds of being published on the platform.

Similar enforcement actions have consistently affected content covering Khalistani separatist groups, regional insurgencies, and related security issues posted by the outlet.

Together, these cases illustrate how account restrictions, reduced algorithmic reach, and eligibility determinations directly shape which arguments reach the broader public and which are suppressed before gaining momentum.

Governance Structures and the Influence of Oversight Bodies

Broader questions regarding Meta’s political neutrality also extend to its governance mechanisms, such as the Meta Oversight Board. Investigations published by OpIndia in 2020 revealed that 18 of the initial 20 board members had professional ties or connections to organizations funded by George Soros’s Open Society Foundations. Additional reports highlighted controversies surrounding individual board members, such as Tawakkol Karman and her past ties with Yemen's Islah party.

These background details raised legitimate questions about whether a body presented as ideologically diverse was primarily selected from a narrow subset of international institutions. These structural choices remain significant given the board's influence over how political speech, national security issues, and regional movements are evaluated globally.

Executive Acknowledgments of External Political Pressures

Meta’s moderation practices have also been influenced by external government communications. In an August 2024 letter addressed to the US House Judiciary Committee, Meta CEO Mark Zuckerberg stated that senior officials from the Biden administration had repeatedly pressured Meta to censor specific COVID-19 content, including humorous and satirical posts. Zuckerberg expressed regret that the company was not more vocal in resisting this pressure at the time, stating that the government demands were improper.

He acknowledged that Meta made enforcement choices during that period that it would not make today with the benefit of hindsight.

This admission confirmed that state actors and political entities actively attempt to exert influence over platform moderation policies, and that social networks have historically adjusted their enforcement practices in response.

The Structural Reality of Algorithmic Amplification

Meta does not require an explicit partnership with the Cockroach Janta Party to significantly amplify its campaign material.

Instead, the overall mechanics of the system—which prioritizes watch time, rewards high-friction visual content, applies complex boundaries between documentation and violence, relies on flawed automated moderation, and operates under documented external pressures—inherently favors dramatic protest footage.

Videos highlighting clashes involving law enforcement spread under the umbrella of political documentation, while opposing material faces stricter content checks. Outraged reactions inadvertently boost engagement metrics, and passive viewing time signals to the algorithm that similar clips should be recommended to more feeds.

While Meta categorizes this process as individualized algorithmic personalization, to an end user receiving unsolicited political material, it functions as systematic narrative amplification. Ultimately, the primary power of major technology platforms lies in their ability to govern reach—determining which political perspectives become unavoidable and which become hidden from view.