When strange and misleading answers to search queries generated by Google’s new AI Overview feature went viral on social media last week, the company issued statements that generally downplayed the idea that the technology had problems. On Thursday night, the company’s head of search, Liz Reid, admitted that the errors had highlighted areas that needed improvement, writing that “we wanted to explain what happened and the steps we have taken.”
Reid’s post directly referenced two of the most viral and wildly incorrect AI Overview results. It was seen supporting Google’s algorithms eating stones because doing so “may be good for you,” and the other suggested using non-toxic glue to thicken pizza sauce.
Stone consumption is not a topic that many people have written about or asked questions about online, so there are not many sources for a search engine to turn to. According to Reid, the AI tool found an article from La Oniona satirical website that was republished by a software company and misinterpreted the information as factual.
As for Google telling its users to put glue on pizza, Reid effectively attributed the mistake to a failure of humor. “We saw AI overviews that featured sarcastic or troll content on discussion forums,” he wrote. “Forums are often a great source of authentic, first-hand information, but in some cases they can lead to unhelpful advice, such as using glue to make cheese stick to pizza.”
It’s probably best not to prepare any kind of AI-generated dinner menu without reading it carefully first.
Reid also suggested that it would be unfair to judge the quality of Google’s new vision for search based on viral screenshots. He stated that the company conducted extensive testing before its launch and that the company’s data shows that people value AI overviews, including indicating that people are more likely to stay on a page discovered that way.
Why the embarrassing failures? Reid characterized the errors that drew attention as the result of an Internet-wide audit that wasn’t always well-intentioned. “There’s nothing like millions of people using this feature with lots of fresh searches. We have also seen new meaningless searches, apparently intended to produce erroneous results.”
Google claims that some widely distributed screenshots of AI overviews gone wrong were fake, which appears to be true based on WIRED’s own testing. For example, a user at X posted a screenshot that seemed to be an overview of AI answering the question “Can a cockroach live on your penis?” with enthusiastic confirmation from the search engine that this is normal. The post has been viewed more than five million times. However, upon closer inspection, the screenshot format does not align with how AI overviews are actually presented to users. WIRED was unable to recreate anything close to that result.
And it’s not just social media users who were fooled by misleading screenshots of fake AI overviews. The New York Times issued a correction to its report on the feature and clarified that AI Overviews never suggested that users should jump off the Golden Gate Bridge if they experience depression; that was just a dark meme on social media. “Others have implied that we got dangerous results on issues like leaving dogs in cars, smoking during pregnancy, and depression,” Reid wrote Thursday. “Those AI overviews never appeared.”
However, Reid’s post also makes it clear that all was not well with the original form of Google’s big new search update. The company made “more than a dozen technical improvements” to AI Overviews, he wrote.
Only four are described: better detection of “meaningless queries” that are not worthy of an AI overview; make the feature less reliant on user-generated content on sites like Reddit; offering AI overviews less frequently in situations where users did not find them useful; and strengthen barriers that disable AI summaries on important topics like health.
There was no mention in Reid’s blog post of significantly rolling back AI summaries. Google says it will continue to monitor user feedback and adjust features as necessary.