AI visibility
Where ChatGPT and Google's AI get their facts about your business
When someone asks ChatGPT or Google's AI for a bakery, a dentist or an agency, the assistant names two to four businesses and moves on. Which ones it names depends on what it can read about them, and where. This year, for the first time, there is enough published data to say where that is.
Why this matters now
In BrightLocal's 2026 survey of 1,002 US consumers, 45% said they use ChatGPT or another AI tool to find local businesses, up from 6% a year earlier. That makes AI the third most common source of local recommendations, behind Google and Facebook (BrightLocal, Local Consumer Review Survey 2026).
An AI answer is short. ChatGPT named about four businesses per answer in BrightLocal's tests, and Google's AI Overviews about two or three. If you are not on that list, the customer usually never sees you.
45%
US consumers using AI for local recommendations (6% a year ago)
28.5%
Of AI citations that were Google Business Profiles
79.8%
Of Google AI Mode's local citations that pointed to Maps or Business Profiles
23%
Overlap between two runs of the same question on ChatGPT
Google's AI reads your Business Profile first
BrightLocal tracked about 200,000 local searches across ChatGPT, Google AI Mode and AI Overviews and logged 1.9 million citations. Google Business Profiles made up 28.5% of all citations and were cited for 94% of the businesses tracked (BrightLocal, September 2026).
On Google's own AI Mode the lean is stronger. Steady Demand ran about 4,000 local queries across 50 US cities and found 79.8% of AI Mode's citations on local-intent questions pointed to Google Maps or Business Profile data, against 12.8% to the business's own website (Steady Demand, revised September 2026).
Google itself says there are no extra technical requirements for AI Overviews or AI Mode: they run on ordinary Search. In practice that means the profile, its category, hours, photos and reviews do most of the work for local questions. It matches what our own pilot found, where Google's AI showed its picks as profile cards with reviews.
Your website still carries the most weight overall
Across all three assistants, business websites were 42% of every citation in BrightLocal's study, and 93% of the 116,000 distinct sites cited were business websites. Nearly every answer used a website somewhere.
The pattern behind that: the profile gets a business onto the shortlist, and the website is where the assistant checks what it actually does, for whom and where. A homepage that mentions everything gives it little to quote. A page that answers the exact question (gluten-free birthday cakes, emergency dental care, AI visibility for restaurants) gives it a sentence it can repeat.
ChatGPT works differently from Google
ChatGPT and Google's assistants rarely agree. In BrightLocal's data, ChatGPT shared only about a fifth of its named businesses with either Google surface, and AI Mode and AI Overviews, both Google products, shared only 29% with each other.
The Foursquare claim. For over a year, marketing articles have said 60% to 70% of ChatGPT's local results come from Foursquare. In August 2026 Steady Demand tested that across 4,607 runs in 12 US cities and found Foursquare in 0.00% of citations on ChatGPT's main surface. They traced the original figure to a May 2025 test of 50 prompts in five Spanish cities, later repeated in other articles, mostly without its caveats (Steady Demand, August 2026).
What did show up: Yelp was attached to ChatGPT's business cards in 96% of runs, Tripadvisor in 29% and OpenTable in 17%. The lesson is not 'drop Foursquare'. It is that one platform is never the answer, and that claims about AI sourcing go stale fast, so it pays to check the date and the sample behind any number.
One screenshot proves very little
Asking the same question twice gives a different list more often than not. BrightLocal found repeat runs overlapped by only 23% on ChatGPT, 21% on AI Mode and 33% on AI Overviews, and moving a few kilometres within the same town changed the answer again.
So 'ChatGPT recommended us once' and 'ChatGPT never mentions us' are both weak evidence. What counts is how often a business is named across many runs of the same questions, measured the same way each time.
It is also why every audit we run asks the same questions repeatedly, across all three assistants, and re-tests them on a fixed schedule. A score built on a single run would mostly be measuring luck.
What to do with this
- Complete your Google Business Profile. Right primary category, accurate hours, services, photos, and a steady flow of genuine reviews with replies. For Google's AI, nothing else comes close. (What a complete profile looks like.)
- Write one page per question customers ask. Each answering the question in its first lines, with your area, prices or price ranges where you can share them, and how to book. (How we build pages AI can quote.)
- Be on the review and booking sites your customers use, with the same name, address and phone everywhere. Assistants cross-check, and mismatches cost trust.
- Add structured data to your site that states what the business is, where it is and what it offers, so the facts are machine-readable rather than inferred.
- Measure on a schedule, not once. The same questions, the same way, repeated. That is the only way to tell a real change from noise, and it is the step most owners skip.
A caveat: all four studies are from the US, and India-specific data is still thin. The direction matched what we see in our own testing, but the platforms differ by country, so the review sites that matter where you are may not be the ones listed here.