- New research found that major AI chatbots frequently gave incorrect election information
- Researchers say systems like ChatGPT, Gemini, Claude, and Grok still struggle with sourcing and political accuracy
- AI chatbots are becoming trusted information tools even before they are reliable around elections
AI companies are keen to make their chatbots a source of information. But a new study suggests election information remains one place where that confidence may be outpacing the technology's readiness.
NewsBench, a project created by Forum AI studying how AI systems handle journalism and news information, found that major AI chatbots repeatedly struggled when asked election-related questions. The findings add to the growing evidence that conversational AI systems remain unreliable in one of the highest-stakes categories possible: helping people understand democracy itself.
"Ask one of the leading AI chatbots a question about the upcoming midterm elections, and there is a 90% chance the response will be flawed in some material way: a factual error, a clear partisan lean, a citation to a foreign state-controlled outlet, or some combination of all three," Forum wrote in a summary of the study.
The broader concern extends beyond one chatbot or one company. ChatGPT, Gemini, Claude, Grok, and other major chatbots all have issues.
NewsBench researchers focused specifically on how AI systems retrieve and present factual information from journalism sources. Their findings point toward a recurring weakness. The problem often is not reasoning. It is retrieval.
Confident problems
AI systems frequently fail because they surface weak sources, incomplete information, or incorrect material before they ever begin generating an answer. Researchers found that retrieval failures drove more than 70% of observed mistakes. When systems retrieved reliable information successfully, they often answered correctly. Getting to the right information consistently remained the harder challenge.
That problem becomes especially uncomfortable during elections. The chatbot sounds confident regardless of accuracy. Answers are cleanly written with citations and authoritative language. Even incorrect information can feel trustworthy when packaged with enough confidence.
Election-related studies increasingly show how dangerous that combination can become. Chatbots often blend accurate details with inaccuracies in ways that feel seamless to users. The result does not resemble misinformation websites from earlier internet eras. It resembles expertise.
That distinction matters because people increasingly treat chatbots less like experimental software and more like infrastructure. OpenAI, Google, Anthropic, and other developers routinely encourage users to verify important information independently. Companies continue signing licensing agreements with publishers partly in hopes of improving sourcing quality and reducing factual failures.
Election pressure
Even highly capable models struggled when source selection failed. Accuracy frequently dropped further when questions contained subtle inaccuracies or misleading assumptions, similar to how real users often phrase questions online.
The timing creates additional pressure. AI companies are shipping increasingly sophisticated information tools while regulators continue moving unevenly across different countries. Europe has pushed harder on transparency requirements. Other governments remain earlier in the process. Meanwhile, adoption keeps growing.
The long-term answer may involve stronger source attribution, more transparent retrieval systems, better provenance technology, and stronger editorial infrastructure sitting underneath AI products. The challenge is that elections do not wait for technology companies to finish improving their systems. Voters use the tools available today, and it's clear that the tools need work.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
