Google AI Mode hallucinates Anne Burrell death

This is the kind of failure that turns an AI answer engine from convenience into liability: the model sounds confident, cites sources, and still leaves you with less truth than you started with. Google says AI Mode won’t always get it right, but when the error touches suicide or death, the UX gap becomes a safety problem, not a nuisance.

• –Google launched AI Mode as an experimental Search layer with Gemini, query fan-out, and follow-up questions, so factual drift is a known risk rather than an edge-case bug.
• –The worst part here is the contradiction loop: the system surfaces links, then tells users not to trust them, which makes the answer feel both authoritative and self-negating.
• –Sensitive queries need a hard abstain-and-link path, especially around death, suicide, and mental health, where conversational polish should never outrun caution.
• –The broader product lesson is trust: if AI Mode can’t beat regular search on factual reliability, users will keep defaulting back to blue links.