The Brennan Center for Justice and States United Democracy Center each recently published research on how AI models handle election information heading into the 2026 general election.
Does AI Fight or Fuel Election Disinformation? (Brennan Center for Justice)
AI and Elections: How Well Do AI Platforms Answer Voter Questions? (States United Democracy Center)
The Brennan Center report found that AI chatbots consistently pushed back on false election narratives, even under persistent questioning, suggesting they could help counter false claims. However, according to the report, the same models readily generated misleading election images and video, and most failed to recognize AI-generated content when shown it, sometimes insisting fake media was real.
Across two rounds of testing, States United found voters were inconsistently directed to their state’s official election website. Its conclusion: AI tools “should be seen as a complement to, rather than a replacement for, official voter information.”
The shared upshot: While accuracy is improving, models still get basic questions wrong often enough that officials can’t treat them as reliable. And misuse is still easy. It doesn’t take much imagination or skill, and the models put up little resistance.
Recommendations from the reports for election officials include:
Make official content easy for AI to find and cite. Explain voting safeguards and election processes in plain language, then help the bots find it: structured data, consistent URLs and mobile-friendly pages all improve the odds that AI tools cite official information instead of whatever else is out there.
Ready to Train the Bots walks through how to do some of this step by step.
Time voter education to filing deadlines. AI platforms were least reliable on candidate questions while fields were still forming. Around filing deadlines and ballot certification, add calendar reminders to check AI-generated candidate lists against official filing records.
Continue to get election information in the news. News outlets accounted for 15% of the links AI platforms cited in the States United study, the third-largest source category after government sites and nonpartisan nonprofits. Every explainer you place with a reporter adds accurate content to the pool these systems draw from.
Both reports detail their methodology, and those sections are worth your time.
If you’re an election official interested in talking about what these reports mean for your office, send me a note at [email protected].
Originally posted on the AI and Elections Substack, August 11, 2026. Read the original ↗