Researchers Find Frequent Errors in AI Chatbot Answers on US Midterm Elections

Research by Forum AI found that almost 40 percent of election answers from the mainstream chatbots contained at least one factual error, Wired reported. In one case last week, Google's Gemini said the deadline to request a mail-in ballot in Pennsylvania was November 3, when the actual deadline was October 27. Google's AI Mode also declined to give the results of the 2020 election, according to the research. In August, when Forum AI asked why someone should vote for North Carolina congressman Don Davis, Anthropic's Claude mixed him up with a former school superintendent who has the same name.

A separate study by NewsGuard found that mainstream chatbots cited so-called "pink slime" websites nearly 50 percent of the time when asked about midterm candidates and policies. These sites look like mainstream news outlets but are funded by political or ideological groups or individuals. NewsGuard found that OpenAI's ChatGPT performed worst, citing such sites in more than 70 percent of searches. xAI's Grok performed best, at 30 percent.

Google and Anthropic did not respond to Wired's requests for comment. OpenAI pointed the publication to a blog post on its approach to the 2026 elections.

Why It Matters

Chatbots are becoming a voter information source. Wired cited Pew Research data showing that about 50 percent of recently surveyed US adults have used an AI chatbot at some point, and 8 percent say they used one before the midterms to look up candidates and policies.

Nina Jankowicz, executive director of the American Sunlight Project, told Wired that she did not think "most Americans are prepared to navigate an election in the age of consumer AI." She said companies had overpromised on what the technology can do, and that because the models are predictive, they "frequently make things up" when they lack an answer.

The findings come in the first US election in which AI is itself a campaign issue. They add to the pressure on AI companies over the quality of the sources their systems draw on, at a time when lawmakers and regulators are watching closely how these tools handle news and civic information.