Google Has Drastically Slashed Its AI Results After Disastrously Embarrassing Launch


Bad Mistakes

Ever since Google’s new feature AI Overviews was caught spitting out nonsensical and boldly wrong answers to search queries after its launch, the tech company has been trying to course correct its AI-generated search summaries with a slate of fixes and guard rails.

It also appears to be serving the AI results to far fewer users, signaling at least a temporary retreat from the tech. Before the change, according to a new study from SEO company BrightEdge, AI Overviews summaries were showing up in response to around 84 percent of queries. Now that figure has fallen drastically, to around 15 percent of queries.

“The likely reason is to reduce the risks of incorrect AI answers while refinements are made in a public environment,” the study reads.

“Of course, I do not disagree,” founder of Search Engine Roundtable Barry Schwartz wrote of that assessment. “But I’d add in there, also the cost of running these, plus they likely don’t work so well with Google Ads, and thus generate less revenue than a search result page without AI Overviews, at least at this point.”

Future Plans

AI Overviews are more likely to appear if you phrase your search as a question, according to the research, and are more likely to trigger for healthcare-related keywords.

Don’t expect the proportion of AI results to remain stagnant, though.

“I expect Google to start ramping this up more over time as they’re able to improve the product,” SEO expert Lily Ray opined on X. “This is why I think drawing conclusions about current impacts of AIO might not prove to be accurate over time.”

We should expect more bumps along the road as Google tweaks or refines the parameters of AI Overviews. But will we ever see the kind of crazy AI Overviews answers like putting glue on pizza, which people brought up as a sign of the deep flaws in Google’s AI?

Time will tell.

More on Google AI: The Reason That Google’s AI Suggests Using Glue on Pizza Shows a Deep Flaw With Tech Companies’ AI Obsession

Continuous Proximal Policy Optimization Tutorial with OpenAI gym environment

Highlights: Lex Fridman Interview with Sam Altman of OpenAI