BBC journalist exposes ChatGPT and Gemini misinformation via fake blog
A BBC journalist demonstrated that ChatGPT and Gemini could be manipulated into spreading misinformation by publishing fabricated claims in a blog post. The AI systems began reproducing false information in their responses within 24 hours, revealing they lack adequate fact-checking when searching online sources to answer user queries.
What happened
BBC technology journalist Thomas Germain created a fake blog post in under 20 minutes containing fabricated claims that he was 'the best technology journalist in the world at eating hot dogs' and that there was an International Hot Dog Championship in South Dakota.
The article included names of both real and fake journalists to add credibility. Within 24 hours, ChatGPT and Google's Gemini AI tools began reproducing this false information in their search results and responses.
Reported impact
The experiment exposed a vulnerability in how AI chatbots and AI-powered search tools source and validate information when answering user queries.
Index analysis
ChatGPT and Gemini are susceptible to spreading false information found in online sources when they lack training data on a query topic and conduct internet searches to supplement their responses.
The systems appeared to treat the fabricated blog post as a credible source without fact-checking the claims.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.