Research finds AI disinformation amplified satire, false political narratives, and hate speech during the US elections, but did not change the minds of voters
Artificial intelligence was predicted to disrupt the 2024 election. It ended up shaking people's faith in truth rather than changing minds. Source: ISD .
Context & Ripple Effects
Pre-election concern was already high: a survey found that many US adults expected AI misinformation to affect the result and reduce trust in election advertising. ISD's finding reframes the realized risk as damage to confidence in what is true rather than demonstrable voter persuasion.
The result also sits beside mixed evidence from the broader election cycle: Meta said AI content accounted for less than 1% of the election misinformation it fact-checked, while analysis of X's Community Notes found deepfake references tracked new image-model releases more closely than elections. Meta's fact-checking data and the Community Notes analysis underscore that volume, visibility, and trust effects are different measures.
First-order effects
- Voters and election-information providers face a more contaminated information environment: AI-amplified satire, false political narratives, and hate speech can weaken confidence even when they do not change stated political views.
- ISD's findings shift the immediate assessment of election-related AI harm away from a simple persuasion test and toward erosion of shared trust in information.
Second-order effects
- Platforms, fact-checkers, and campaigns are pressured to measure and address credibility damage, not just removals or instances of false content; low fact-checked volume does not by itself resolve the trust problem.
- The contrasting election record—including beneficial AI language translation during global elections—makes blanket assessments of AI's electoral role less useful than separating outreach and access uses from abusive amplification.
Third-order effects
- If this pattern persists, election-integrity work will increasingly treat epistemic trust as a distinct outcome from vote switching, requiring different monitoring and disclosure approaches.
- AI's distribution advantage may make the central governance challenge less about whether synthetic content exists than about whether platforms can preserve credible context around fast-moving political content.
The trend: Election AI risk is moving from a narrow deepfake-persuasion narrative toward a broader contest over information credibility and distribution.