You rank on Google Maps. ChatGPT still names someone else.

Your name is on the map. It is not in the answer.

Picture a homeowner in Langley. Her heat pump quit during a November rainstorm. She does not scroll a page of blue links. She asks ChatGPT, “Who can fix a heat pump tonight?” She gets three or four names, each with a short reason.

She calls the first one. That job is booked. Your Google profile has ninety reviews and sits in the top three on the map. You were not in her answer. Nothing in your reports will show that call, because it never reached you.

Until now, nobody had measured how often this happens. In September 2026, a large study did.

What the study measured

BrightLocal ran 200,875 local searches. None used a business name. They were things a customer would type, like “emergency dentist open late tonight.” Each search ran on ChatGPT, Google AI Mode and Google AI Overviews. It ran from nine points on a map around each of 1,342 business locations, over 60 days.

The answers used about 1.97 million sources, called citations. BrightLocal then counted who got named and where the facts came from.

One honest note. BrightLocal sells an AI tracking tool, and it ran this study with that tool. We checked the numbers against its published pages. The limits are in the section near the end.

Maps names you twice as often as AI does

Google Maps showed a tracked business in 66% of searches. The AI tools did far worse.

Share of searches where a tracked business was named. Source: BrightLocal, 2026.

Maps is also steadier. BrightLocal measured 99% agreement across repeat attempts on Maps. On the AI tools it was 91% to 96%.

Why the gap? Maps runs one local ranking system. An AI tool works another way. Google calls it “query fan-out.” Here is how it goes:

Google says this itself. Its guide to AI features says AI Overviews and AI Mode “may use a ‘query fan-out’ technique” and that there are “no additional technical requirements” to appear. The basics of good SEO still apply.

Maps ranks you for one search. AI runs several.

Three AI tools, three different answers

You might guess the AI tools agree with each other. They do not. Google’s two AI tools shared only 29% of the businesses they named. ChatGPT matched either one just 19% to 20% of the time.

Same town. Same question. Different list. The reason is where each tool reads. Pick a tool to see its main sources.

Shares are our arithmetic from BrightLocal’s tables. Each is a share of that tool’s citations to outside sites, not counting a business’s own website.

So a business can win on Maps, show up in AI Overviews, vanish in AI Mode and never appear in ChatGPT.

Checking one tool tells you almost nothing about the other three.

Your website is the part you control

One number from this study got passed around: 93%. It is real, but it is easy to misread. Flip the switch to see both ways to count.

Both counts say the same thing. AI reads business websites at huge scale. In our experience, many local sites are a home page, an About page and one Services page. That gives an AI very little to match against a question like “Who fixes a heat pump in Langley tonight?”

BrightLocal also points to a report from SEO expert Lily Ray. ChatGPT now runs “site:” searches in its research. That means it looks inside a business’s own website for the answer.

Distance still counts, too. Between 53% and 59% of the businesses the AI named were within 5 km of the searcher. ChatGPT reached much further. 90% of its picks fell within 287 km. For Google’s tools the figure was 39 to 64 km.

Storefronts got named a bit more than service-area businesses: 33.6% against 29.4%. If you drive to your customers, you have no storefront to lean on. Your pages have to carry the proof. Our guide to the service-area setting shows how.

Trade by trade: where the AI looks

BrightLocal also split its citations by type of business. Pick yours. These samples are small, so read them as a pattern, not a promise.

Share of that trade’s citations. Bars are scaled to a 60% maximum so the small ones stay visible. Source: BrightLocal, Aug 2026.

Two things stand out. Google Business Profile is the biggest single source in every trade. Yelp is second in all four. After that, the list depends on the trade: Angi and the BBB for general services, Zocdoc for dentists, Justia and Avvo for lawyers.

One catch for Canadians. BrightLocal’s locations were in the US, UK and Australia. Canadian directories were not measured. Use this as a starting list. Then ask the AI tools about your own trade in your own city and see which sites they cite.

Find the questions AI is already asking about you

Here is a free trick from the same BrightLocal study. Google Search Console already lists the long, chatty searches your site appears for. They sound a lot like the way people talk to AI. You only need one filter.

Try the patterns here first. Pick one, then paste in some of your own queries. Nothing leaves your browser.

Want longer or shorter searches? Change the number in the curly braces. In ^([^ ]+ ){7,}[^ ]+$ the 7 means “seven words each followed by a space, then one last word,” so it finds searches of 8 words or more. Use {3,} for 4 or more, or {11,} for 12 or more. The tester above does the math for you in the custom box.

We ran it on our own data

Check it by hand: the scorecard

BrightLocal found that one check is not a measurement. Run the same search twice and the two lists share only 21% to 33% of their businesses on average. A business that showed up once came back in about half of the later tries. That was 50% on ChatGPT and AI Mode, and 58% on AI Overviews.

So check more than once. Write down the five questions your best customers ask. Ask each tool, on three different days. Count how often you are named. Then fill in your counts.

What to do about it

Most of what works for Maps also works for AI. The difference is that AI reads more places. Here is the short list.

What this study can’t tell you

A good source is one that admits its limits. So here are the limits we found.

• It is vendor research. BrightLocal sells the AI tracker that ran the study. That does not make the numbers wrong. It means you should know who gains from you caring about the topic.
• It is not a Canadian sample. The companion directory study used locations in the US, UK and Australia. Treat Canadian results as untested.
• The trade samples are small. The plumber, dentist and lawyer tables rest on roughly 20 to 34 locations each.
• The page has a few typos. It gives 200,875 searches in the methodology and 200,085 in the text. It gives Yelp’s share as 9.53% in the table and 8.53% in the text. We used the methodology figure and the table figure.
• “Named” is not “chosen.” The study counts who was named. It does not count who got the call.

We do not guarantee that any AI tool will name your business. Nobody honest can. We take one client per trade, per city. What we can do is make sure your site, your profiles and your proof give the AI something real to find.