Within less than 1 week of appearing to have doubled the AI overview in search results, Google actually appears to be shrinking this feature.  According to a Wired report citing figures from SEO firm BrightEdge, AI overviews only appear in 11% of search queries, down from 27% when it first launched last month.

Obviously your AI tools are not very good looking to tell people to eat rocks and put glue on their pizzas. But while the company certainly hasn't given up on the overview feature, the last thing it needs is to mock that it's the main way users are exposed to the tool. So it makes sense that you might want to dominate the AI overview and tweak what you actually suggest to people.

BrightEdge claims it has been tracking an overview of AI since it launched as an opt-in feature for beta testers in 12 months last year. Since then, I have figured out the type of query that generates an Overviews response, and apparently it seems most likely to appear in a healthcare-related search query. According to the report, that's 63% of the time.

Obviously, it is rather a concern, especially if AI summaries tend to provide information that is not accurate. Medical-related searches already lead people to the worst possible conclusion, it searches for symptoms and suddenly your sore throat that you have something more serious.

While e-commerce-related searches clearly had a 23%-time summary, ai summaries did not "rarely" appear in restaurant and travel-related searches. . Misinformation can be irritating, but in general the kind of thing that is not so serious. Actively promoting misinformation on medical outcomes can put people's lives at risk.

The fact that an AI overview can pull information from a lot of sources, including things like Reddit and random blog posts, is also relevant. Overviews adds a list of sources, but it doesn't explicitly tell you which information is coming from where. Everyone has their own prejudices and agendas, and if you don't know where the information is coming from, it's much harder to figure out if you can trust it or not.

After all, you can't necessarily trust someone who might be trying to sell you something. Similarly, you can't trust that a random Reddit post is 100% accurate to see who is writing it or how well known they are.

That means this is something Google needs to keep on top of, even after the AI overview has stopped encouraging people to eat glued heavy pizza sauce. This means that you will not be able to access the site without the permission of the user. 

A Google spokesperson told Wired that the company "continues to improve the timing and method of displaying an AI overview, so it is as useful as possible, including technical updates from the past 1 week to improve the quality of responses."We also declined to share official figures on how often AI summaries appear in search results, but noted that BrightEdge's figures do not reflect what we are seeing internally.

In the meantime, be sure to take the AI summary results obtained with a pinch of salt. This is also what you need to do with regular Google search results anyway.