The Dangers of Relying on AI Chatbots for Election Information

As voters head to the polls, an increasing number of Americans are turning to artificial intelligence-powered chatbots to help inform their electoral decisions.

Recent data from Pew Research indicates that approximately 50 percent of U.S. adults have utilized AI chatbots, with 8 percent specifically using these tools to research candidates and policy positions ahead of the midterm elections.

However, relying on these tools poses significant risks, as chatbot outputs are frequently outdated, hallucinated, or derived from highly partisan sources. Even prominent, ostensibly non-partisan models—including OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini—have demonstrated notable deficiencies regarding midterm election queries.

A recent study by Forum AI highlighted these inaccuracies; for instance, Google’s AI Mode declined to provide 2020 election results, and Gemini incorrectly stated the Pennsylvania mail-in ballot deadline as November 3, missing the actual October 27 cutoff. The Forum AI research revealed that nearly 40 percent of election-related answers provided by mainstream chatbots contained at least one factual error. In one instance in August, Claude incorrectly identified North Carolina Congressman Don Davis by confusing him with a former school superintendent of the same name.

While Google and Anthropic did not respond to requests for comment, OpenAI directed inquiries to a blog post detailing its strategy for the 2026 elections.

A primary driver of this misinformation is the chatbots’ dependence on a compromised information ecosystem. Research from NewsGuard shows that mainstream chatbots cite “pink slime” websites—platforms that masquerade as legitimate news outlets but are actually funded by ideological or political interest groups—nearly 50 percent of the time when queried about midterm candidates and policies. This phenomenon has intensified leading up to the elections. According to NewsGuard, ChatGPT was the most frequent offender, citing “pink slime” sites in over 70 percent of searches, whereas xAI’s Grok performed relatively better, citing such sites in 30 percent of cases.

“I don’t think that most Americans are prepared to navigate an election in the age of consumer AI,” Nina Jankowicz, executive director of the American Sunlight Project and a former disinformation official during the Biden administration, told WIRED. “One of the problems is that executives and companies have really overpromised on what AI is capable of. It has to ingest this information from somewhere, and often those sources are not very good. And it’s predictive, so if it doesn’t know the answer, it does frequently make things up.”

Source link