06/05/2026
I asked Google's AI Mode about Colombian football players on a Friday night.
It gave me a correct answer.
I pushed back on one detail.
It panicked.
It threw out EVERYTHING. Including the correct parts.
Then it told me Luis Dรญaz was still at Liverpool. He's been at Bayern Munich all season. He assisted Kane's goal in the DFB-Pokal final two weeks ago.
So I tested it: "So Dรญaz and Kane haven't played together?"
Gemini's response: "I completely panicked and contradicted myself. I am so sorry."
The AI wasn't corrected by facts. It was corrected by tone.
Now think about this: organizations are making million-dollar decisions based on AI outputs. Strategy. Hiring. Investments. Market research.
All on systems that can be talked out of a correct answer by skepticism.
This isn't a Gemini problem. ChatGPT, Claude, Perplexity. They all share the same limitation.
Every time you accept an AI answer without verifying it, you're voting. You're telling these companies that accuracy doesn't matter.
Your vote matters. Be critical. Verify. Push back.
Full breakdown on the Navigamo blog. https://www.navigamo.co/post/hallucination-by-gaslight-lucho-diaz-beats-gemini
A simple question about Colombian football players broke Google's AI Mode. Gemini hallucinated, contradicted itself, and named its own failure: Hallucination by Gaslight. Here's what that means for SEO and GEO.