Your business panel is now an AI Overview

A fused business panel is a Google result where the local knowledge panel for a business carries an "AI Overview" label and a machine-written description, so the first thing a searcher reads about you is Google's summary rather than your hours, phone number and photos.

Google has not announced this. It was spotted on mobile and documented by Search Engine Roundtable on September 17, 2026, after Ben Fisher posted it on LinkedIn and Barry Schwartz reproduced it. The panel shows the AI Overview label at the top and a "Show more" control that opens a chat-style interface for follow-up questions. Schwartz noted that the description of his own company drew on outdated content from his corporate website.

That last detail is the part worth acting on. A knowledge panel used to be a structured record you could mostly control through Google Business Profile. A fused panel is a generated summary, and its inputs are whatever Google decides to read about you.

#What the first field study found

Fisher followed up with a field study of 100 local businesses across 10 categories, a mix of national chains and single-location firms, searched from New York (Steady Demand, September 26, 2026). Each business was searched four ways: the name alone, "[name] reviews", "is [name] legit" and "what does [name] do".

The headline results:

• 26 of 100 businesses showed an AI Overview from the business-name search alone on mobile. On desktop the same searches produced one for 6 (Steady Demand).
• Among businesses that triggered an AI Overview, 80% showed the fused version, where the overview sits inside the business listing itself (Steady Demand).
• 99 of the 100 businesses triggered an AI Overview on at least one of the four searches (Steady Demand).

The study is explicit that these are single-day snapshots, with noticeable variation within hours. Treat the percentages as a description of one test on one day, not a stable rate. The direction is still clear: on mobile, a branded search for a local business now has a real chance of opening with a generated description, and the modifier queries people type before they call or buy almost always produce one.

#Your website is not the main input

The more useful finding is about sources. Fisher also checked where AI Mode got its information when it answered questions about these businesses.

Among businesses that already ranked in Google's normal top 10 for their own name, 74% were still not cited by AI Mode when it answered about them, meaning the business's own website was not used as a source (Steady Demand). Across the four searches, social and review sites outranked the business's own website about 59% of the time.

The most cited sources in the study were Yelp (118 citations), Google reviews and Maps (109), Reddit (91), Trustpilot (77) and the Better Business Bureau (39) (Steady Demand).

The pattern has a plain explanation. Branded and evaluative queries ("reviews", "is it legit") pull the engine toward third-party evidence, because that is the evidence that answers the question. A homepage that says the company is trustworthy is not an answer to "is this company legit". A review profile with several hundred ratings is.

#Wrong business, not wrong detail

The failure that should worry operators most is entity confusion. In four cases the study found the AI Overview described a different business entirely: searches for Fox Pest Control returned Fox Wildlife Services, Paragon Sports returned Paragon Basketball Academy, Sobo & Sobo returned Sobo Pools, and Chapel Hill Tire returned Big Brand Tire & Service (Steady Demand).

Four out of 100 is a small count from one snapshot. It is still a qualitatively different error from an outdated opening time. A classic knowledge panel is keyed to a verified Business Profile. A generated summary is keyed to the engine's reading of a name, and names collide. Businesses with generic or shared names, or names that overlap with a larger company in another city, carry the highest risk.

#What to do this week

None of this needs new tooling. It needs a few checks that most local teams have never run because the panel used to be predictable.

Search yourself the way customers do. Run your business name on a phone, logged out, along with "reviews", "is [name] legit" and "what does [name] do". Fisher recommends the same set and also suggests searching from a location where you do not operate, which is the easiest way to spot name collisions (Steady Demand). Repeat it more than once, since results moved within hours.

Fix errors at the source, not in the panel. If the summary is wrong, find the page it is paraphrasing. Schwartz's example came from his own outdated website copy (Search Engine Roundtable). In other cases it will be a stale Yelp category, an old Reddit thread or a BBB record with a former address. Editing your Business Profile through Google's own tools is still necessary, but it no longer controls the whole description.

Audit third-party profiles as owned assets. Given that Yelp, Google reviews, Reddit, Trustpilot and the BBB led the citation counts, each claimed profile should carry the same name, category, address, service list and description as your site. Unclaimed or abandoned profiles are now inputs to the first sentence a customer reads.

Make the "what do you do" answer explicit on your site. One plain paragraph near the top of the homepage or about page that states what the business is, where it operates and what it does not do. This gives the engine a clean definitional sentence and reduces the chance it borrows one from a similarly named company. Google's guidance on AI features in Search says the same fundamentals apply as for ordinary search, so there is no special markup to add for this.

Remove stale pages. Old location pages, retired service pages and archived press releases still get read. If a page describes something that is no longer true, update it or remove it.

#How to track it

Rank tracking will not show this. Your name can sit at position one while the panel above it describes you using a review site, or describes someone else. The useful measurement is the generated text itself and its sources: for each branded query, record whether an AI Overview appears, whether it is fused into the panel, which URLs it cites, and whether it names the right business.

Because the study showed variation within hours, a single check is not enough. Sample the same queries on a schedule, keep the cited URLs, and treat a change in the source mix as the signal. When a review site replaces your own domain as the main citation, that is the moment to look at what the review site says.

This is still a test, and Google may change or drop it. The underlying shift is unlikely to reverse: the summary of a local business is moving from a record the owner maintains to text the engine writes from whatever it can find.