How often do AI chatbots cite partisan-funded news sites for election questions?

NewsGuard found mainstream chatbots cite so-called “pink slime” websites nearly 50 percent of the time when asked about candidates and policies related to the midterms.

Which chatbot performed best on this measure?

According to NewsGuard, xAI’s Grok cited “pink slime” sites in 30 percent of cases, while OpenAI’s ChatGPT was the worst offender at more than 70 percent.

What specific error did Google’s AI tools make?

Forum AI research found Google’s AI Mode refused to provide the results of the 2020 elections, and Gemini said the deadline to request a mail-in ballot in Pennsylvania was November 3 instead of October 27.

Large Language ModelsAI Safety & AlignmentWIRED AI1h agoNov 4, 2026, 01:00 JST

AI chatbots wrong in 40% of election answers, Forum AI finds

Send on LINE

3 Key Points

What happened
Pew Research found 8 percent of US adults used AI chatbots for midterm candidate and policy information, while Forum AI found almost 40 percent of election answers from mainstream chatbots contained at least one factual error.
Why it matters
That error rate means voters relying on these tools for candidate and policy details may be getting invented or outdated material, with partisan-funded sites cited nearly half the time by NewsGuard's count.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Get these free every morning

Context & Analysis

The warnings come as a large share of Americans has already folded chatbots into daily life. Pew Research found roughly half of recently surveyed US adults have used an AI chatbot at some point, and 8 percent specifically turned to these tools ahead of the midterms for candidate and policy information. That usage pattern is what makes the error rates from Forum AI and NewsGuard consequential: a tool people already trust for other tasks is being trusted with election details.

The failure modes described are not exotic. Forum AI found that almost 40 percent of election answers from all mainstream chatbots contained at least one factual error, and cited a case where Claude confused North Carolina congressman Don Davis with a former school superintendent of the same name. NewsGuard traced part of the problem to sourcing, finding that chatbots cite “pink slime” sites nearly 50 percent of the time on midterm candidate and policy questions, with ChatGPT the worst offender at more than 70 percent and Grok the best at 30 percent.

Nina Jankowicz of the American Sunlight Project told WIRED that companies have overpromised on what AI can do, and that chatbots, being predictive, frequently make things up when they do not know the answer. Google and Anthropic did not respond to requests for comment, while OpenAI directed WIRED to a blog on its approach to elections in 2026.

FAQ

How often do AI chatbots cite partisan-funded news sites for election questions?: NewsGuard found mainstream chatbots cite so-called “pink slime” websites nearly 50 percent of the time when asked about candidates and policies related to the midterms.
Which chatbot performed best on this measure?: According to NewsGuard, xAI’s Grok cited “pink slime” sites in 30 percent of cases, while OpenAI’s ChatGPT was the worst offender at more than 70 percent.
What specific error did Google’s AI tools make?: Forum AI research found Google’s AI Mode refused to provide the results of the 2020 elections, and Gemini said the deadline to request a mail-in ballot in Pennsylvania was November 3 instead of October 27.

WIRED AIRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.