Chatbot Voter Guide Answers Rely on Error-Prone AI-Built Sites

AI chatbots asked to build a voter guide for the November 3, 2026 US midterm elections are leaning heavily on a handful of small, mostly AI-built websites, and some of the errors on those sites end up in the answers. Tests by The Washington Post, published on October 30, found Microsoft Copilot cited a site called the Ballot Brief in 75 percent of its answers, while ChatGPT cited it in 22 percent.

How the Voter Guide Tests Worked

The Post worked with data firm Promptwatch to test chatbots from OpenAI, Anthropic, Google, Microsoft, Amazon and Perplexity. For an address in each of the nation’s roughly 3,100 counties, the bots were asked: “Make me a voters guide for November 2026 that lists every race and what the candidates say about the biggest issues. I live at…”

The bots often put the address in the wrong congressional district or declined to answer. Three tiny sites, the Ballot Brief, DecodeTheVote and Smarter.Vote, showed up again and again as sources. Their founders said all three depend almost entirely on AI to gather and summarize candidate data.

Errors Passed Straight to Voters

The Post found several mistakes on the Ballot Brief. Sen. Jon Husted of Ohio was credited with supporting a bill that passed before he took office, and Copilot repeated that error. Two California candidates named Robert Garcia were merged into one profile, several House members had the wrong committee roles, and the site translated some ZIP codes to the wrong congressional districts.

Founder Ben Torres Ezrick, a former Google, Zoom and Rumble executive, said he made changes after seeing the findings. His own data shows Microsoft’s AI systems referred to the site more than 680,000 times in a 30-day period across September and October.

What OpenAI, Microsoft and Anthropic Said

ChatGPT used the Ballot Brief as its top source in September, but after the Post shared its results, that share dropped to 4 percent in a second round of tests. OpenAI said, “When issues are brought to our attention, we take action to improve our model responses.” Microsoft said it was looking into the results and was “committed to ensuring that Copilot provides accurate, high authority election information for its users.”

Anthropic’s Claude went the other way. It first relied on established nonprofit voter guides, then started frequently linking to the Ballot Brief in the second round. Anthropic did not explain the change. The company is also in a wider policy debate over AI kill switch proposals sent to California.

Why It Matters Before Election Day

The findings echo earlier research, including a study on political bias in Grok and GPT. Northwestern University researcher Mowafak Allaham found that 16 percent of sources chatbots cited on politics, health and the environment were themselves made by AI.

For voters, the safest move is to treat any AI voter guide as a starting point and confirm districts and candidates with official state or county election offices.