The New Battlefield of Truth: How AI is Rewriting India’s Protest Narratives

October has arrived with a sobering reminder that the digital landscape we navigate is no longer what it once seemed. Across India, the tools of generative artificial intelligence—once celebrated as marvels of creative potential—have quietly transformed into instruments of confusion, manipulation, and political warfare. For those of us who dedicate our days to verifying what is real and what is fabricated, the challenge has never been more daunting. The lines between political commentary, satire, and outright misinformation have blurred to the point where even experienced fact-checkers find themselves squinting at screens, questioning their own eyes. What we witnessed this past month was not merely a collection of isolated incidents, but rather a deepening pattern—a glimpse into a future where our perception of reality itself can be engineered, weaponized, and deployed against us in ways that feel both alien and unsettlingly familiar.

The second wave of Gen Z-led protests that swept through India, beginning on October 2nd, brought with it a fresh torrent of misleading content that revealed the sophistication of those who seek to shape public opinion through deception. Organized by the Cockroach Janta Party—a name that combines irreverence with defiance—and various student groups, these demonstrations demanded the resignation of Chief Election Commissioner Gyanesh Kumar over concerns about the Special Intensive Revision of electoral rolls. This followed an earlier phase of activism during June and July, when young Indians had taken to the streets over examination irregularities and called for the removal of then Union Education Minister Dharmendra Pradhan. But what made this second phase particularly unsettling was not just the passion of the protestors, but the calculated ways in which their movement was being undermined from the shadows. Fact-checkers at NewsMeter began noticing disturbing patterns: AI-generated personas of women, fabricated visuals involving people connected to the demonstrations, and sophisticated uses of technology designed not to make arguments, but to ridicule, sexualize, and distract.

Perhaps the most troubling development was the circulation of AI-generated female personas in fabricated protest scenes—videos depicting women in revealing outfits against protest-related backdrops, creating the illusion that these were genuine participants in the demonstrations. This was not merely about fabricating presence; it was about introducing a sexualized visual narrative that hijacked online conversations, steering them away from the substantive issues driving the protests and toward lewd commentary about women’s clothing and appearance. The strategy was insidious precisely because it did not make an explicit political claim—it simply shifted attention, polluted discourse, and diminished the visibility of women’s genuine participation in political movements. Who was creating these personas? What narratives were they promoting? These questions remain unanswered, but the implications are clear: synthetic imagery has become a tool for undermining the credibility and dignity of activists without ever engaging with their arguments.

The political weaponization of generative AI did not stop at the protest movement. With Uttar Pradesh approaching Assembly elections in early 2027, a new form of political rivalry has emerged—one that uses gender-swapped videos to mock leaders like Chief Minister Yogi Adityanath and Samajwadi Party leader Akhilesh Yadav. These videos, showing the leaders in sarees performing exaggerated dance sequences, relied on familiar gender stereotypes to reduce political opponents to objects of ridicule. The strategy is clever in its simplicity: instead of confronting a leader’s policies or public statements, the content attacks their appearance, using gender expression itself as a weapon of mockery. This represents another dimension of AI-generated political content that demands our attention. Where does legitimate satire end and manipulative content begin? How do we distinguish between creative political commentary and the deliberate use of gender stereotypes to demean public figures? These are questions that will only become more pressing as election cycles intensify and the tools of content creation become more accessible to those who would use them for less than noble purposes.

The limitations of our verification tools have also been thrown into sharp relief this month, particularly through an investigation into photographs shared by Delhi Police following the October 15th protests. What we discovered was both alarming and instructive: authentic photographs, after undergoing minor modifications or enhancements, were being flagged as AI-generated by detection tools. The implications are profound. If genuine photographs can be dismissed as fabricated because of minor edits, then trust in authentic imagery is itself under threat. Detection scores alone, we learned, cannot establish whether an image is authentic or synthetic. Even seemingly insignificant modifications can influence results dramatically. The original source, visual details, and different versions of an image must all be examined before any judgment is made. This finding exposes a critical vulnerability in the verification process: in our eagerness to combat AI-generated misinformation, we risk becoming so reliant on automated tools that we fail to exercise the human judgment that remains essential to determining truth.

This brings us to the urgent need for transparent AI labelling and content provenance standards. Organizations and individuals sharing AI-assisted visuals must clearly disclose whether AI was used and explain the purpose of the intervention. There is a vital distinction between an entirely AI-generated image and an authentic photograph that has undergone AI-assisted enhancement—yet without proper disclosure, even genuine photographs can be dismissed as fabricated. Companies are developing invisible watermarking technologies and content provenance standards such as C2PA, with tools like Google’s SynthID designed to identify AI-generated or AI-modified content. Even minor edits made using generative AI features can leave traces. However, as our investigations have shown, the presence of an invisible watermark does not automatically establish that an image was entirely generated by AI. It indicates AI involvement but does not explain what was changed or how extensively the original was modified. Clear labelling and explanations of editing processes are therefore essential—not just for verification teams, but for audiences who must interpret such content in their daily lives.

For those seeking to sharpen their investigative skills in this increasingly complex landscape, open-source intelligence—OSINT—offers a path forward, though it demands practice, experimentation, and continuous learning. Understanding how to apply tools in different investigative situations makes all the difference between superficial knowledge and genuine expertise. Practical challenges, Capture the Flag exercises, and investigations conducted by experienced practitioners provide invaluable opportunities to test techniques and discover new approaches. Five platforms stand out as particularly useful for those eager to explore: the Bellingcat Challenge, OSINT Industries, the UK OSINT Community, Sofia Santos’ Gralhix, and Michael’s OSINT CTF. Each offers unique opportunities to develop investigative confidence, question assumptions, and approach digital evidence more critically. As the field evolves, there is always something new to learn—and for those committed to the pursuit of truth in an age of synthetic media, this continuous growth is not just a professional necessity, but a civic duty. The struggle to distinguish reality from fabrication is no longer confined to fact-checkers and journalists; it is a challenge we all must confront, and the skills we develop today will shape the integrity of our public discourse tomorrow.