Website Poisons Chatbots With Fabricated Claims

A security blogger published a fabricated article on his personal website on Feb. 25, 2026, and found that within 24 hours major chatbots repeated the false claims. Google’s Gemini and ChatGPT echoed the bogus ranking while Anthropic’s Claude resisted, highlighting how web-indexed models can ingest unverified content and risk dataset poisoning for downstream LLM outputs.