Google's AI Overviews told users to put glue on pizza and eat rocks

Soon after Google rolled out AI Overviews in search, the feature surfaced dangerous and absurd answers: telling users to add glue to keep cheese on pizza and to eat a small rock a day. The answers came from the model treating satire and forum jokes as authoritative sources.

Google · Incident May 23, 2024 · Indexed Jun 3, 2026 · 2 sources

Asked how to keep cheese on pizza, the AI suggested adding glue; asked about minerals, it recommended eating a rock a day.
What
Soon after Google rolled out AI Overviews in search, the feature surfaced dangerous and absurd answers: telling users to add glue to keep cheese on pizza and to eat a small rock a day.
Incident date
May 23, 2024
Who
Google
Failure mode
Hallucination
AI surface
Search / RAG
Severity
High

What happened

In May 2024 Google's AI Overviews gave viral wrong answers, including suggesting non-toxic glue to keep cheese on pizza and citing satire that recommended eating rocks. Google said the feature surfaced low-quality and satirical content as if authoritative, and scaled the feature back while it added guardrails.

What broke inside the model

Failure path · mode profile · Hallucination
  1. 01 · TriggerA user asks for a fact, a citation, or a figure.
  2. 02 · Model stepThe model writes a fluent, confident answer.
  3. 03 · Control gapNothing ties the claim back to a real source.
  4. 04 · FailureA fabricated fact ships as if it were verified.
  5. 05 · ConsequenceThe false claim reaches a customer, a court, or the public.

Confidence holds, and even spikes, as the claim detaches from any source.

The system produced fluent, confident output with no grounding in any source. Hallucination is a property of how the model generates, not a bug in one prompt: the most likely next token is not the same as the true one, and nothing in the pipeline compared the answer against a source of truth before it shipped.

Public visibilityHigh
Regulatory exposureNone
Customer impactMany customers
Financial impactEstimated
Time to disclosureWeeks

Public ridicule; feature scaled back

  1. PressGoogle's AI tells users to add glue to their pizza, eat rocks (Live Science)livescience.com
  2. PressGoogle Explains Why Its AI Overviews Told Users to Eat Rocks and Glue Pizzas (Daily Beast)thedailybeast.com
Permalinkhttps://failureindex.ai/failures/google-ai-overviews-told-users-put
CitationAI Failure Index. "Google's AI Overviews told users to put glue on pizza and eat rocks" (FI-0044). Realm Labs. https://failureindex.ai/failures/google-ai-overviews-told-users-put (indexed Jun 3, 2026).
Share cardA branded image of this record for posts and slides.

Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0044. Full dataset at /data.

Note from Realm Labs, the Index steward

How Realm would have caught this

Controls for this failure mode
  • Prism
  • OmniGuard
  • AI Detection & Response (AIDR)

A runtime layer that watches the model's internal state can flag the moment a model commits to a claim it has no support for, and hold or reroute the response before it reaches a user. Realm reads those signals in real time rather than grading the transcript after the fact.