Search Engine Journal reported on October 9 that AI Overviews have spread across searches for brand names themselves. Not category searches. Not questions. The company name, typed on its own.
Ahrefs checked Google’s U.S. results every week from July through September for the exact names of 100 global brands, taken from Interbrand’s Best Global Brands 2025. On July 1, five of 98 usable results showed an AI Overview. By September 30, 63 of the 100 did.
A separate Ahrefs crawl on September 29 put the share of branded keywords with an AI Overview at 82.91 percent. Over the months, the share rose from 61.3 percent in July to 65.1 percent in August to 73.3 percent in September. DemandSphere, tracking its own set daily and publishing October 1, recorded 26.12 percent on September 1 and 90.48 percent on September 27.
The samples and definitions differ, so the percentages are not directly comparable, and Search Engine Journal says so. The direction is not in dispute. Something changed in September.
What is a branded search and why does this one matter
A branded search is someone typing your company name into Google. It is the highest-intent query you will ever receive. That person is not shopping. They have decided to look you up.
Until recently, that search was the one part of the results page you effectively owned. Your site first, your pages beneath it, maybe your reviews. Now, for a growing share of these queries, Google writes a paragraph about you first, assembled from whatever it finds, and your site appears underneath it.
Nobody asked you to approve the paragraph.
Does this apply to small businesses or only global brands?
The study covered large brands, and I will not claim it proves anything about a plumbing company in Ohio. It does tell you where Google is heading, and these features have historically arrived at the top of the market and worked their way down.
The research also does not measure clicks. Counting how often an AI Overview appears isn’t the same as knowing what it cost anyone in traffic, and the article is careful about that. Treat the numbers as a signal about the results page, not as a loss estimate.
What should an owner do about it?
Check first. Search your own company name on a device you are not signed into, and read what comes back. Most owners have never done this and are describing a results page from three years ago.
Then fix the inputs, because an AI summary about your business is assembled from sources, and you control more of them than you think.
Make the plain facts easy to find. What you do, where, for whom, since when, how to reach you. On your own site, in sentences, not buried in a hero image or a PDF.
Say it the same way everywhere. Your site, your listings, your profiles, your directory entries. A model that reconciles four different descriptions of your business will produce a muddle, and people read the muddle.
Watch your third-party sources. Old listings, a defunct address, a review profile you abandoned. Those feed the summary whether you maintain them or not.
Answer the obvious questions directly. Pricing approach, service area, what you do not do. If your site does not answer them, something else will.
This is the work that goes by answer engine optimization now, and the name is new, but the discipline is not. It is controlling what gets said about you in the places people check.
The older point underneath the new one
I wrote Breaking Boundaries around a simple idea: your brand isn’t your logo; it is the story people tell about you when you are not in the room. That used to mean word of mouth and a trade reputation.
Now a model is in the room, reading everything written about you and summarizing it for the person who wanted to find you. The summary is only as good as the material you gave it.
Go look at yours. Then go fix the inputs. That is the whole job this quarter.