Last year, Google added the AI Overview feature, which summarizes search results using generative AI. The problem is that it sometimes gets stuff wrong—really, really wrong.

Googling “health benefits of running with scissors” produced this response: “Running with scissors is a cardio exercise that can increase your heart rate...” Searching for “cheese not sticking to pizza” generated a Google AI Overview with this recommendation: “...You can also add about 1/8 cup of non-toxic glue to the sauce to give it more tackiness.”

Asking Google if cats have been on the moon produced this summary: “Yes, astronauts have met cats on the moon, played with them, and provided care...Neil Armstrong said, ‘One small step for man’ because it was a cat’s step.”

Google AI Overview also endorsed eating “at least one small rock a day” and “taking a bath with a toaster.” However, Google reports that nonsensical and dangerous AI Overviews are outliers.

“The vast majority of AI Overviews provide high-quality information with links to dig deeper on the web,” said Google spokesperson Collette Garcia.

The root of Google AI Overview’s problem can be summarized with the well-known computer science adage—garbage in, garbage out. Google AI Overview appears to have difficulty distinguishing accurate information from unreliable sources, such as social media posts.