The AI Hallucination Registry
A public, timestamped record of false claims AI engines make about real brands — every entry backed by the engine's logged answer and the official source that contradicts it.
What is the AI Hallucination Registry
The AI Hallucination Registry is a public database of machine-detected false claims that AI engines such as ChatGPT, Gemini, and Claude make about real brands. Each entry is timestamped and paired with the engine's logged answer and a quoted official source that contradicts it.
Live incidents
No incidents published yet. Monitoring is live and runs nightly; verified incidents will appear here as they're detected.
Join the watchlist to be alerted the moment your brand is affected.
Join the watchlistHow it works
We ask
We ask engines verifiable questions about brands.
We verify
Answers are compared against quoted snippets from official pages.
We publish
Only mechanically verified contradictions are published — ambiguous answers are never shown.
Watch your brand
Get an email the moment a verified incident involving your brand is published.
Frequently asked questions
What is a brand hallucination?
A brand hallucination is when an AI engine states something false about a company — a wrong founding year, headquarters, price, or a claim with no basis in the company's own published information.
How are incidents detected and verified?
Each question is asked to multiple AI engines from different model families, multiple times per night. Two independent AI judges compare every answer against a quoted snippet from the company's official page. An incident is only published when both judges agree it's a genuine contradiction, the quotes are verified to be real, and the same false claim recurs across repeated checks — not a one-off fluke.
My brand appears — how do I correct it?
Contact us with the incident URL and the correct, sourced information. Verified corrections are resolved and removed from the live registry.
How do I check my brand?
Join the watchlist below to be alerted the moment a verified incident involving your brand is published. Full self-serve search is coming soon.
What does "still live" mean?
An incident stays marked as live for as long as our nightly checks keep reproducing the same false claim from the engine. Once the engine stops repeating it or the underlying issue is corrected, the incident is resolved and no longer shown as live.
Source: AI Hallucination Registry, GEOscanAI. Retrieved Jul 21, 2026. geoscanai.co/registry