What I tested

I picked 100 real local businesses across ten categories, including plumbers, dentists, barbershops, movers, tire shops, gyms, and restaurants.

There was a mix of national chains and single-location businesses.

For each business I searched:

• the business name
• “[business name] reviews”
• “is [business name] legit”
• “what does [business name] do”

The searches were done on mobile from New York on the same day.

One search per business, per query. That’s important.

The number I care about: 26%

Before running the study, I picked two things I really wanted to measure.

• How often an AI Overview appeared when someone searched only the business name.
• If an AI Overview appeared on any of the four searches.

The first one ended up being much more interesting.

26 out of 100 businesses showed an AI Overview from the business-name search alone.

That’s a very normal search. Someone hears about a plumber. They Google the company.

Now there’s a decent chance AI is part of what they see.

The other number was 99%.

99 of the 100 businesses triggered an AI Overview on at least one of the four searches.

I don’t think that number is nearly as useful.

Ask Google about a company four different ways and eventually you’re probably going to trigger an AI Overview.

The 26% is the number I’d pay attention to.

Then I found the fused layout

Of the businesses that triggered an AI Overview, 80% showed what I’ve been calling a fused result.

“Fused” is my term, not Google’s.

Instead of having an AI Overview and then a separate business listing, Google combines them.

You see:

• Business name.
• Rating.
• Hours.
• Photos.
• Call.
• Directions.
• Website.
• And the AI-generated information.

All in the same block.

That’s your Google Business Profile fused with the AI answer.

And I think that’s a pretty big change.

A bad address, wrong hours or other bad information in your Business Profile isn’t just making your listing look wrong anymore. Even hallucinations.

It can make Google’s AI answer about your business look wrong too.

Mobile and desktop were very different

This was another thing that stood out.

Same business names. Big difference.

Side note: Google will usually test on mobile before publishing to desktop, so this is normal.

So if you’ve been checking your own business from your laptop and saying, “I don’t see any of this,” check your phone.

There may also be a location problem

I noticed something else while doing the manual checks.

Businesses with several nearby locations seemed less likely to get an AI Overview than businesses that resolved cleanly to one obvious location. (For example, Walmart or another chain-designated search.)

You can rank for your own name and AI can still ignore you

I also looked at AI Mode. Just to be clear, AI Mode and AI Overviews aren’t the same thing.

An AI Overview is the AI answer Google puts into normal search results.

AI Mode is Google’s separate conversational search experience.

I wanted to see where AI Mode was getting information about these businesses.

And this is where it gets interesting for Local SEO.

Among businesses that already ranked in Google’s normal top 10 for their own name, 74% still weren’t cited by AI Mode when it answered about them.

Google knows exactly which business you are. You rank for your own name.

And AI Mode still goes somewhere else for information about you.

Across the four searches, social and review sites outranked the business’s own website about 59% of the time.

The biggest sources were:

The practical part is pretty simple.

Go look at what Yelp, Reddit and Trustpilot say about your business. Especially if someone is asking Google something like: “Is this business legit?”

Your website isn’t necessarily the source Google wants for that answer.

Then I checked the API against a real phone

Most of this data came from an API.

That’s normal. I’m not manually running hundreds of searches every time I do one of these studies.

But I wanted to know whether the API’s classification actually matched what a person would see.

So I started checking the results on my phone.

Brand by brand.

And it wasn’t perfect.

These are small samples. I wouldn’t do much with the exact percentages yet.

But the order is useful.

“No AI Overview” was the API’s most reliable call.

“Plain AI Overview” was the worst.

And digging into those misses is where I found the strangest part of this whole study.

Google showed me the wrong business

Four times. Not a slightly wrong description. Not an old phone number.

The wrong business.

• Fox Pest Control→Fox Wildlife Services
• Paragon Sports→Paragon Basketball Academy
• Sobo & Sobo→Sobo Pools
• Chapel Hill Tire→Big Brand Tire & Service

Chapel Hill Tire gave me Big Brand Tire & Service, which isn’t just the wrong business. It’s a competing tire chain.

There was no warning. Google didn’t say it wasn’t sure. It just answered.

This tracks with how Gemini can be overconfident.

I dug deeper into Fox Pest Control because I wanted to understand what happened.

Fox Pest Control didn’t have a location near where I was searching.

So Google appears to have found another nearby business with a similar name and decided that’s what I meant.

For the other three, I haven’t verified the cause yet. They may be the same issue. They may not.

But the result is the same: I searched the exact name of one business and Google confidently told me about another one.

That’s worth testing yourself.

Especially if your business has a name that’s similar to another company, a common phrase, or something that exists in several categories.

Search your business name from somewhere you don’t have a location. See who Google thinks you are.

Then the results started changing

This was the other problem.

I ran some of the same searches again a few hours later. They changed.

One business went from having no AI Overview at all to a full fused result.

I followed one national moving company across three days.

1. Day one: Fused.
2. Day two: Nothing.
3. Day three: Plain AI Overview.

Same company. Same query. Three different results.

I also did a smaller three-day repeat test.

Only 1 out of 20 brands changed layout day to day in that particular sample, so I’m not saying everything is constantly flipping.

But it happens. And that’s enough to change how I think these results should be measured.

Phoenix is next

I also collected the same 100 businesses from Phoenix because I wanted to see how much location changes the result.

That work isn’t finished yet.

Only 38 of the 100 businesses actually had locations in both New York and Phoenix.

Of those, 34 showed nothing in either city.

Only three produced a different outcome between the two locations.

So right now? I don’t have much there.

Maybe location matters. Maybe it matters only under certain conditions.

I need more data before I’m going to tell you either one.

This isn’t finished