Mainstream artificial intelligence chatbots are providing inaccurate information to voters, with research showing nearly 40 percent of election-related responses contain factual errors or rely on unreliable partisan sources.
Key Points
- Pew Research indicates 8 percent of U.S. adults have used AI chatbots to research candidates and policies ahead of the current election cycle.
- Forum AI found that Google’s Gemini and other mainstream tools frequently hallucinate data, including incorrect mail-in ballot deadlines and conflated candidate biographies.
- NewsGuard reports that chatbots often cite "pink slime" websites—partisan outlets disguised as news—in up to 70 percent of election-related queries.
- OpenAI’s ChatGPT cited these unreliable sources in 70 percent of searches, while xAI’s Grok performed best with a 30 percent citation rate.
- Major AI companies, including Google and Anthropic, have faced criticism for failing to provide accurate, nonpartisan information regarding election logistics and candidate records.