Viral AI Images as Real Events
AI-generated images, video, and audio depicting fabricated events are regularly shared on social media as genuine documentation of real occurrences. This is a confirmed, documented phenomenon that affects political discourse, crisis communication, and public trust. UNESCO, the Brennan Center, and numerous fact-checking institutions have documented the harms.
What we know
Generative AI tools capable of producing highly realistic images, video, and audio have become widely accessible since 2022–2023. These tools allow individuals with no specialist knowledge to create convincing synthetic media in seconds. Viral spread of AI-generated images falsely representing real events has been documented across multiple contexts: political campaigns, armed conflicts, natural disasters, and celebrity impersonation.
Documented examples include: AI-generated images of former US President Trump being arrested (2023) that circulated before his actual arrest; AI-generated audio of Slovakia's Progressive Slovakia party leader discussing election fraud, released days before the 2023 election; a fake video of Ukrainian President Zelensky ordering soldiers to surrender, broadcast via a hacked TV station in 2022; AI images from hurricane disaster zones misrepresenting the scale of destruction; and thousands of AI-generated images circulating during the 2024 US presidential election. In the financial sphere, a deepfake video call impersonating the CFO of Arup engineering resulted in a $25 million transfer in 2024.
A complicating factor is the 'liar's dividend': public awareness that deepfakes exist allows individuals to falsely claim that authentic, incriminating content is AI-generated. This cuts in both directions, eroding trust in genuine evidence as well as enabling fabricated content.
Unesco describes deepfakes as creating a 'crisis of knowing.' Research from the University of Waterloo found that only 61% of participants could distinguish AI-generated faces from real ones. Newsrooms and fact-checkers report increasing difficulty as the technology improves. Detection tools exist but are in an adversarial race with generation tools. The Brennan Center for Justice has documented the threat deepfakes pose to electoral integrity.
Common claims
- AI-generated images are obviously fake and easy to spot.Increasingly false. Modern AI generation produces images that a majority of people cannot distinguish from photographs. Research shows detection rates close to chance for high-quality outputs.
- If an image goes viral on a credible platform it must be real.False. Platform moderation cannot stop viral spread of AI-generated content before fact-checkers can react. Multiple well-documented cases show AI images reaching millions of views before removal.
- This video must be AI-generated because it shows something implausible.Context-dependent. Implausibility alone is not a reliable indicator; authentic footage of unusual events can look 'AI-like', and the 'liar's dividend' means real content can be falsely dismissed.
Evidence hierarchy
All sources
- Deepfakes and the crisis of knowingUNESCO · 2025
- Deepfakes, Elections, and Shrinking the Liar's DividendBrennan Center for Justice · 2024
- The apocalypse that wasn't: AI was everywhere in 2024's electionsHarvard Kennedy School Ash Center · 2024
- AI-generated disinformation poses threat of misleading voters in 2024 electionPBS NewsHour / Associated Press · 2023