AI chatbots now nail basic voting facts almost perfectly — so why do they still often fail to point you to the right official source?

This explores how accurate AI chatbots are on election topics, and whether factual accuracy is the right measure at all. The corpus gives two different answers depending on the kind of question asked. On narrow voter-logistics questions, accuracy has improved a lot. States United found that ChatGPT and Google's AI had reached a 0% rate of verifiable factual errors by early 2026 Do AI chatbots give voters accurate election information?. The same study found weaker spots, though. The chatbots sent voters to official state election websites less than half the time, gave incomplete candidate information, and leaned on sources anyone can edit, such as Wikipedia. Being correct and being a dependable source turned out to be different things.

On broader news questions, the picture is much worse. Journalists from 22 public broadcasters in 18 countries reviewed more than 3,000 AI answers about news. Nearly half had significant errors, a third had serious sourcing problems, and a fifth had major accuracy failures, including made-up facts Do AI assistants reliably answer questions about news?. Election coverage sits in this territory too: debates, candidate claims and fast-moving events. One way to read the two studies together is that chatbots do well on stable facts like registration deadlines and much worse once a question needs current events or interpretation.

The most surprising finding is that a chatbot can make no factual errors and still distort an election. When the Dutch Data Protection Authority tested chatbots as voting advisers, they recommended the same two parties, PVV and GroenLinks–PvdA, in over 56% of cases, even when a user's stated positions matched a different party Do chatbots steer Dutch voters toward the same parties?. Centrist parties came up less than 2% of the time. The regulator called this a 'vacuum cleaner effect' that pulls voters toward the extremes Do AI chatbots systematically bias voters toward extreme parties?. It traced the cause to chatbots drawing on unstructured internet text rather than structured data on party positions. None of this would appear in a fact-check. The bias lies in which answers the chatbot gives, not in any false statement.

There is also a gap between how accurate chatbots are and how much people trust them. In a 199-person study, people's existing trust in AI predicted whether they checked a chatbot's claims better than whether the answer was actually right did. A warm, friendly tone also made people more likely to agree with wrong answers Does access to web search prevent overreliance on chatbots?. Related work finds that trust follows how expert a chatbot sounds, not how accurate it is Does chatbot language style actually shape how much we trust it?. So better accuracy doesn't automatically mean better-informed voters.

For now, the stakes are limited by how few people use chatbots this way. Only 15% of U.S. voters plan to use chatbots for election information, compared with 68% who plan to use news articles, and most say they doubt the chatbots are accurate Will voters actually use AI chatbots for election information?. Worldwide, weekly chatbot use for news is around 10% and growing, especially among people under 35. Trust in chatbot answers lags well behind trust in news overall Why do AI chatbots gain news users but lose their trust?, and among people who do use chatbots, trust is what drives that use Does trust in AI chatbots drive news-seeking behavior?. When chatbots do correct false beliefs, the effect can be real: short conversations with a chatbot representing the other party reduced partisan misperceptions, although most of the gains faded within a week Can AI chatbots reduce partisan misperceptions and warm cross-party feelings?.