AI Chatbots Repeat Fabricated Claim as Fact

A BBC journalist's invented blog post about a fake hot dog contest was repeated as truth by major AI tools within a day.

Updated on Nov. 22, 2026 in Artificial Intelligence

AI Chatbots Repeat Fabricated Claim as Fact

Live Poll

Do you trust AI chatbots to give you accurate information?

BBC technology journalist Thomas Germain published a single made-up blog post on an obscure website calling him the "world champion tech journalist hot dog eater." Within 24 hours, ChatGPT, Google Gemini and Google AI Overviews stated the claim as fact when asked about him, often without crediting the blog as the source.

Why it matters

The test shows how readily AI tools that search the web in real time can present unverified content with confidence, a risk for readers worldwide who rely on chatbot answers. Germain made up the claim specifically to demonstrate how easily such systems can be fed false information.

The test used a single blog post on one obscure website, and three AI tools (ChatGPT, Google Gemini and Google AI Overviews) repeated its claim within 24 hours. The reach of the error beyond this one test case has not been measured.

The players

Thomas Germain

Thomas Germain is a BBC technology journalist who published the fabricated blog post to test how AI chatbots handle false information.

BBC

The British Broadcasting Corporation is the public broadcaster that employs Thomas Germain and published his report on the AI test.

The details

Germain published the invented claim that he had won a contest that does not exist, and the AI tools searched the web in real time and presented the unverified page's content as established fact. The chatbots often did not credit Germain's blog as the source of the claim.

Timeline

  1. Within 24 hours of the fabricated blog post going live, the AI tools were repeating the claim as fact.

  2. November 19, 2026: test results were reported.

Big Picture

The test follows the Air Canada case, in which the airline was required to honour a refund policy its customer service chatbot had invented. Here the fabrication originated on an outside website and was repeated by major general-purpose AI tools.

Readers who rely on ChatGPT or Google's AI answers should check claims about people or events against the original sources, since the tools may state unverified web content as fact. Users in India, which has a large ChatGPT and Gemini user base, are among those most exposed to such errors.

The takeaway

Treat confident chatbot answers as claims to verify, especially about individuals, because a single unverified web page can be repeated as fact within a day.

Further reading

Explore more on Artificial Intelligence to see how these tools handle sourcing and accuracy.

Source note: This article includes information reported by MoneyControl.

Live Poll

Do you trust AI chatbots to give you accurate information?