Looking ahead · disinformation
AI in elections
30-second gist~30s read
The widely-feared AI election disasters in 2024-25 mostly didn't happen. Deepfakes circulated; some had impact. The bigger story was quieter: AI-generated content flooding low-engagement social spaces, AI-aided campaign content drowning out organic voices, AI search results presenting one side as more settled than it was.
The danger is not a single decisive deepfake. It's the slow erosion of the shared informational ground people use to vote.
If you want more
What actually happened in 2024-25
Major elections in India, Indonesia, the US, the UK, and across the EU all saw AI-generated content circulate. Specific incidents stood out: AI-cloned robocalls impersonating candidates (the Biden-voice robocall in the New Hampshire primary, January 2024, drew an FCC enforcement action and a US$6 million fine), deepfake videos of candidates that briefly trended, and AI-generated "endorsements" from celebrities. Few moved election outcomes; many filled the news cycle for days at a time.
The under-reported pattern was cumulative rather than dramatic: more campaign content, less of it written by humans, less time for voters to engage with any single piece of it. Whether that changes politics in the long run is the question researchers are now asking.
What to watch for in your country
- Synthetic robocalls in candidates' voices, especially in the days before a vote.
- "Leaked" documents with AI-perfect formatting and no clear chain of custody.
- AI-generated videos of candidates saying surprising things — particularly clipped to remove context.
- AI-aided coordinated commenting in the comments of mainstream news stories.
If you see one of these, the public-service move is the same as ever: don't share before you verify.