Pizza glue? Gasoline spaghetti? Google explains what happened to its amazing AI search results

If you’ve been on social media in the past week, you’ve probably seen them. Screenshots of Google’s new AI-powered search summaries have gone viral, mainly because Google allegedly makes bold recommendations like adding glue to your pizza, cooking noodles spaghetti with gasoline or the suggestion that you should eat ice for optimal health.

That’s just the beginning.

Also: How to Avoid AI Overview in Google Search: Three Easy Ways

Other particularly egregious examples have also gone viral, seemingly suggesting a rogue AI feature mix bleach and vinegar to clean washing machines, which can generate potentially deadly chlorine gas, or jump off the Golden Gate Bridge to answer the question “I feel depressed.”

So what happened and why did Google AI Overview recommend it?

First, Google says, Most of what is spread is not true.

Many of the screenshots are simply fake: “Some of these fake results are obvious and silly. Others imply that we returned dangerous results for topics like leaving dogs in cars, smoking while pregnant and depression.” Google says those AI Overviews never appeared.

Second, many of the screenshots are from people intending to get silly search results — like those about eating ice. “Before these screenshots went viral, practically no one asked Google that question,” Google said. If no one is searching for a certain topic on Google, it probably means there isn’t much information about that topic or there is a lack of data. In such cases, it is only the satirical content that the AI ​​understands as accurate.

Also: 7 ways to supercharge your Google searches with AI

Google admitted that some strange or inaccurate results had appeared. Even those questions were for unusual queries, but they revealed some areas for improvement. The company was able to identify the pattern of things going wrong and make more than a dozen technical improvements, including:

• Better detection of meaningless queries that don’t show AI overview and limit the inclusion of satire and humor

• Limit the use of user-generated content in responses that may give misleading advice

• Enable restrictions on queries where AI Overview does not prove helpful

• Don’t show AI Overview for important news topics where freshness and reality are important, and for most health topics

With billions of queries coming in every day, Google says, sometimes things will get strange. The company said it is learning from its mistakes and promised to continue its efforts to strengthen AI Overview.