Key Points
- Public Knowledge said prediction markets, AI chatbots and social-media platforms shaped information during the 2026 United States midterms.
- Researchers found 13.9% of chatbot answers about voting logistics were outdated or inaccurate, while 8% of Americans used chatbots for election information.
- NewsGuard reported that X accounted for 57% of confirmed falsehoods it tracked, while Community Notes failed to flag 93.1% of examined false claims.
- An AI-generated video falsely accused Texas Democratic Senate candidate James Talarico of sexual abuse.
- Public Knowledge called for stronger platform transparency, content-provenance standards and investment in local journalism.
Washington (Public Knowledge) November 18, 2026 – Digital platforms delivered some useful election services but failed to contain significant misinformation during the 2026 United States midterms, according to Public Knowledge’s assessment of prediction markets, AI chatbots and social-media networks. The report said platforms faced greater responsibility after federal election-information capacity was reduced, while voters increasingly encountered political information through automated systems and online communities.
- Key Points
- How reliable were AI chatbots for election information?
- How did AI-generated election content affect candidates?
- What did social-media platforms get right and wrong?
- What was the background to the platform assessment?
- How could these findings affect the audience of What Platforms Got Right and Wrong in the 2026 Midterms?
The assessment found that prediction markets became more prominent after Kalshi’s 2024 legal victory cleared the way for election betting. Kalshi and Polymarket odds were incorporated into mainstream news coverage, although Public Knowledge said no confirmed market-manipulation cases emerged during the midterms. It recommended that platforms publish enforcement data and trading records for independent scrutiny.
How reliable were AI chatbots for election information?
Public Knowledge reported that 13.9% of answers tested in October on voting logistics were outdated or inaccurate. Only 8% of Americans used chatbots for election information during the election, but the report said that figure is likely to rise.
The report identified pink-slime websites, which present partisan or AI-generated material as local news, as a major vulnerability. NewsGuard found that seven leading AI tools cited such sources in nearly half of its tested responses. Anthropic and OpenAI separately entered licensing arrangements with the Associated Press to provide real-time election information and results.
A separate Data and Democracy Initiative assessment of ChatGPT, Claude and Gemini found that 77% of 234 answers about selected races were accurate and complete. Researchers identified 37 partly wrong or incomplete answers, including 22 in Spanish, and found that outdated information was twice as common in Spanish-language responses.
How did AI-generated election content affect candidates?
Public Knowledge reported that a Russian influence operation used AI to create a fabricated video in which a young man falsely accused James Talarico of sexual abuse. The material was quickly debunked, but copies continued circulating online.
OpenAI established an election war room and added safeguards that defaulted some election-related image prompts to cartoon-style outputs. The Brennan Center found that Google Gemini and Meta’s AI were more willing to generate convincing fake election images, while most tested chatbots could not reliably identify AI-generated audio, video or images.
Other reporting described additional AI-generated campaign advertisements and fabricated material involving candidates, showing that synthetic media was used both for political attacks and satire. Public Knowledge said the effect of deepfakes depended heavily on how widely social-media platforms allowed them to spread.
What did social-media platforms get right and wrong?
Public Knowledge said Meta operated an Election Operations Center, directed users towards official voting information, banned some voter-suppression and premature-victory advertisements, restricted new political advertising during the final campaign week and introduced disclosure requirements for some AI-generated content. However, Meta had removed third-party fact-checking and relied on Community Notes instead.
X removed initial posts containing the fabricated Talarico video and suspended users who shared them. Public Knowledge said the first video nevertheless received more than 850,000 views in 24 hours, while copies continued circulating.
NewsGuard reported that X represented 57% of the confirmed midterm falsehoods it tracked and that Elon Musk promoted at least four of them. Its analysis also found that Community Notes failed to flag 93.1% of the false claims examined, including claims promoted by Musk.
TikTok, Meta’s services and X remained major venues for election discussion, while the report said reduced moderation across platforms made it harder to stop false claims before they gained reach. Public Knowledge recommended greater disclosure of moderation performance, stronger provenance systems and better cooperation with independent researchers.
What was the background to the platform assessment?
Public Knowledge had previously examined technology’s role in the 2020, 2022 and 2024 elections, when social media and AI-generated content were central concerns. By 2026, prediction markets and AI chatbots had become additional sources of political information. The assessment also followed a broader retreat from aggressive content moderation after platforms faced allegations that election-related moderation suppressed conservative speech. Public Knowledge said reduced federal coordination and the decline of local newsrooms left platforms with a larger role in supplying and verifying election information.
How could these findings affect the audience of What Platforms Got Right and Wrong in the 2026 Midterms?
Readers of What Platforms Got Right and Wrong in the 2026 Midterms may face more election information through chatbots, prediction-market odds and social-media feeds in future campaigns. The findings suggest that audiences should verify voting logistics with official election authorities and treat automated answers, betting odds and viral media as information requiring independent confirmation. Platforms that publish enforcement data, label synthetic content and support credible local reporting could improve public trust. Without those measures, inaccurate chatbot responses and rapidly shared deepfakes could remain influential even when they are later disproved.
