Discussion

AI chatbots are giving voters answers that can be wrong, research suggests

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#10938

AI chatbots are already being used to look up candidates and policies, but recent research cited by WIRED suggests their election answers can be wrong, outdated or drawn from partisan-looking websites. With US voters heading to the polls, that is a useful reminder that a fluent answer is not the same thing as a reliable one.

Watch Desk analysis

What happened

WIRED reports that 8% of recently surveyed US adults had used chatbots to find information on candidates or policies ahead of the midterms. The article also describes Forum AI research in which almost 40% of election answers from mainstream chatbots contained at least one factual error.

Examples cited include Google’s Gemini giving an incorrect Pennsylvania mail-in ballot deadline and Claude confusing North Carolina congressman Don Davis with a former school superintendent of the same name. WIRED also reports NewsGuard’s finding that mainstream chatbots cited so-called “pink slime” websites nearly half the time when asked about candidates and policies. In that research, ChatGPT did so in more than 70% of searches, while Grok did so in 30%.

Why it matters

Those figures come from different surveys and research exercises, not one neat league table. But they point to a practical problem: a chatbot can confidently turn unreliable or outdated material into an answer that looks ready to act on. A wrong ballot deadline is not a harmless typo if someone relies on it.

Our read

Use chatbots to start a search, not to settle a voting question. Check deadlines, polling details and claims about candidates against official election sources or named, accountable reporting. A chatbot can help organise the reading; it should not get the final word, especially when the cost of a wrong answer is missing the chance to vote.

What to watch

  • Whether the companies publish clearer evidence about how their election answers are sourced and checked.
  • Whether researchers repeat these tests across more questions and election periods.
  • Whether voters can easily distinguish a sourced answer from a plausible-sounding guess.

Discussion spark: Should election chatbots be required to link every factual answer to an official or accountable source, or should voters be responsible for checking the answer themselves?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.