… is producing a flood of bizarre results, including recommendations to eat rocks and use glue on pizza. Screenshots circulating widely on social media show the AI-generated summaries pulling from satirical articles and discussion forums, treating them as factual advice.

Google’s search VP Elizabeth Reid acknowledged the issues in a blog post, noting that while some viral screenshots are faked, “some odd, inaccurate or unhelpful AI Overviews certainly did show up.” The company said it has made more than a dozen technical improvements, including better detection of nonsensical queries and limiting user-generated content in responses. Reid also said the AI Overviews “generally don’t hallucinate” but can misinterpret queries or language on the web.

The company emphasized that content policy violations occurred on “less than one in every 7 million unique queries” and that it is “grateful for the ongoing feedback.”