Local AI visibility is whether a business is named, linked or recommended when someone asks ChatGPT, Google AI Overviews or AI Mode a question with a place in it, such as "can you recommend a dentist in Turin?". In our test on 27 October 2026, covering 9 searches in 3 Italian cities, ChatGPT named 32 businesses: only 6 were also in Google's local pack, but 20 appeared somewhere on Google's first page.
Put simply: ranking first on Google Maps does not put you in ChatGPT's answer.
And it is not only a shopfront problem.
When we asked ChatGPT whether Sunscape is a reliable company, the answer rested on 9 links. One was Google Maps. Sunscape develops solar and agrivoltaic projects and has no walk-in customers, yet its Google listing still ended up among the sources (the full test is in our piece on how AI describes your brand).
What if you have five branches? Or fifty?
Then the problem multiplies by five, or fifty. The three mistakes we see most often:
- measuring only head office, with a generic question and no town in it;
- assuming the local pack is enough, because "ChatGPT uses Google anyway";
- one "find us" page listing every address, instead of one page per location.
In this article:
- the test on dentists, accountants and gyms in Milan, Rome and Turin, answer by answer;
- where ChatGPT gets local businesses from, according to OpenAI and to the links we counted;
- how to measure each location separately, with tools and prices checked today;
- what the chains that appear in all three cities do.
What local AI visibility is (and why it is not the local pack)
Local AI visibility measures whether a business gets named, linked or recommended when the question contains a place: a town, a neighbourhood, "near me". It is one part of AI visibility, with one extra variable: where the person asking is.
The local pack is something else. It is the map with three or four listings that Google shows above the results, ranked on relevance, distance and prominence (we cover it in our guide to local SEO).
AI answers draw on the same sources, but they choose differently.
BrightLocal measured it in a study published on 16 September 2026, across more than 200,000 localised prompts: a business appeared in 66% of Google Maps checks, against 38% for AI Overviews, 33% for ChatGPT and 32% for AI Mode. The study does not state which countries the sample covers, and its examples are American (an emergency dentist in Brooklyn, a pizzeria in Manhattan).
We wanted to see the same mechanism on Italian searches, so we ran our own test.
The test: 9 searches in 3 cities on ChatGPT and Google (27 October 2026)
The method is simple and repeatable. Three categories (dentist, accountant, gym) in three Italian cities (Milan, Rome, Turin). On ChatGPT we asked, in Italian, "can you recommend a [category] in [city]?" with web search on and location set to Italy, through DataForSEO. On Google we read the mobile first page for "[category] [city]", localised to that city. Then we matched the names.
| Search | Businesses named by ChatGPT | Also in the local pack | Also in the AI Overview | Anywhere on Google's first page |
|---|---|---|---|---|
| Dentist, Milan | 3 | 0 | 1 | 2 |
| Dentist, Rome | 4 | 1 | n/a | 1 |
| Dentist, Turin | 4 | 0 | n/a | 2 |
| Accountant, Milan | 4 | 1 | 2 | 4 |
| Accountant, Rome | 4 | 2 | 1 | 3 |
| Accountant, Turin | 3 | 0 | 2 | 3 |
| Gym, Milan | 3 | 0 | 1 | 1 |
| Gym, Rome | 4 | 1 | 0 | 3 |
| Gym, Turin | 3 | 1 | 1 | 1 |
| Total | 32 | 6 (19%) | 8 of 24 (33%) | 20 (62%) |
Three things stand out.
There was an AI Overview on all 9 searches. On the Milan and Rome accountant searches it sat above the local pack.
The local pack and ChatGPT overlapped on 6 businesses out of 32. For dentists in Turin, the first local pack result had 20 reviews; ChatGPT's four picks had between 195 and 1,795.
But 20 of the 32 were still on Google's first page, often in the "related websites" carousel under the map. ChatGPT does not choose outside Google. It chooses a different slice of the same page.
Is one run enough?
No. BrightLocal found that when the same prompt is repeated, only 20% to 33% of businesses are shared between two answers. Our test is a snapshot of one day, as we explain in our test on how accurate AI visibility data is. It shows the mechanism, not a ranking.
The top local pack result can be invisible to ChatGPT.
In our test, 26 of the 32 businesses ChatGPT named were not in Google's local pack. We repeat your customers' questions every month, city by city and branch by branch, and tell you which listing, page or directory is keeping you out of the answer.
Book a strategy callWhere ChatGPT gets local businesses from (listings, directories and websites)
ChatGPT builds a local answer from three kinds of source: business listing data from third parties, sector directories and the businesses' own websites. OpenAI's help page on searching with ChatGPT says that for local businesses a general location may be shared with third-party business listing providers, that answers on iOS and Android can include a map, and that restaurant results can use OpenTable, Resy or Yelp data.
Our test shows it clearly:
- 8 answers out of 9 had a block of structured listings, with address, phone, rating and number of reviews;
- 5 out of 9 linked straight to Google Maps for one or more businesses;
- 6 out of 9 cited directories and portals, such as MioDottore, Wellhub and Tripadvisor;
- the accountant answers for Milan and Rome mostly linked to the firms' own websites.
The international data points the same way. In BrightLocal's study the Google Business Profile is the most cited single source: 28.6% of almost 2 million citations, present for 94% of businesses. Yelp follows with 9.5%. Business websites, taken together, account for 42%.
And AI Overviews?
Even more tied to listings. On the Milan accountant search the AI Overview had 24 references: 20 were Google Business Profiles (one belonged to the professional body itself); the other 4 were the professional body's website, two firms' websites and a pricing guide. If your Google Business Profile is incomplete, so is the raw material of the answer.
You do not need every directory, only the ones that actually show up for your category. In Italy that meant MioDottore and Doctolib for dentists and membership comparison sites for gyms; for the UK picture, see our article on which local citations matter.
Reviews: how many the recommended businesses had (thresholds seen in the test)
In the test, every dentist and accountant ChatGPT recommended had at least 4.8 stars. No exceptions across 22 businesses.
On review counts the gap with the local pack exists, but it is smaller than we expected:
- dentists: a median of 553 reviews for ChatGPT's 11 picks, 461 for the 12 local pack listings;
- accountants: a median of 72 for ChatGPT's 11 picks, 54.5 for the 12 local pack listings.
So you just need more reviews?
No. In Rome ChatGPT recommended a gym with 2 Google reviews; Tripadvisor was among the sources cited next to it. In the Turin local pack, meanwhile, there was an accountant with a single review.
Reviews work as a filter, not a score. Below 4.8 stars, in crowded categories like the ones we tested, you sit outside the group ChatGPT picks from. Above it, other sources decide. How to collect them without pushing is in our guide on how to get Google reviews.
My business has several locations: which platform can measure each one on ChatGPT? (by budget)
A platform that measures each location on ChatGPT has to ask the same question with different towns, or from different positions, and keep a separate history per location. Today there are two ways to do it: a local rank tracker built for Google Maps that has added AI, or a prompt tracker where each location gets its own questions.
The honest answer is that there is no single "AI visibility" number for a multi-location business. There are as many measures as there are towns.
| Approach | Example | How it counts locations | AI platforms | Starting price |
|---|---|---|---|---|
| Local tracker with an AI module | BrightLocal, Track plan | In bands: 1, 2 to 5, 6 to 10 locations | ChatGPT and "Google's AIs" | $41 a month for 1 location, $62 for 2-5 |
| Prompt tracker | ScryRank, Starter plan | A set of questions per town inside the prompt allowance | ChatGPT, Claude, Gemini, Perplexity, AI Overviews | EUR 39 a month for 50 prompts |
ScryRank is the monitoring tool built by Visilay, so we will say it upfront: its public page lists no feature dedicated to locations. In practice a business with 5 branches sets 10 questions per town ("accountant for limited companies in Leeds", "accountant for limited companies in York"...) and fills the 50 prompts of the entry plan. BrightLocal is built for local listings and prices per location, but covers only ChatGPT and Google.
Which one fits depends on where the problem starts. If branches are losing customers on Maps and you want ChatGPT too, the local tracker. If you want to see how AI describes the brand in each market, with the cited sources, the prompt tracker. The other criteria are in our guide on how to choose an AI visibility tool, and full prices for ten tools in AI visibility tool pricing.
How many questions per town and how often to repeat them is covered in prompt tracking. The practical rule for branches: same question, same wording, only the town changes. Otherwise you are comparing two different things. AI Mode has its own rules, covered in how to track Google AI Mode.
What AI sees of a chain: FitActive and Anytime Fitness in 3 cities (one branch at a time)
Gyms are the category that best explains multi-location businesses. In the Milan, Rome and Turin local packs 9 listings out of 12 belonged to chains. In ChatGPT's answers, 5 out of 10.
FitActive, an Italian gym chain, was the only name in all three ChatGPT answers. Never as a generic brand: each time a specific branch, Milan Sismondi, Rome Cipro, Turin Giordano Bruno. For Milan, ChatGPT linked the branch page on FitActive's website, not the home page.
AI Overviews do the opposite. They group by chain and then list branches: on "palestra milano" (gym Milan) FitActive appeared with 4 addresses and Anytime Fitness with 3, and the references included the Google Business Profiles of 5 FitActive and 6 Anytime branches. On the Turin search, Anytime was cited through the pages of 3 different branches on its website.
The branches that made it had two things in common: their own Google listing and their own page, with that branch's address, opening hours and services. The same logic holds outside fitness: the fourth local pack result for dentists in Turin was a practice with three sites.
Google handles chains explicitly. According to Google Business Profile Help, a business with 10 or more locations can request bulk verification and manage its profiles from one spreadsheet. Service-area businesses are not eligible.
How to improve AI visibility for each location (6 steps, cheapest first)
For AI Overviews and AI Mode, Google's page AI features and your website says there are no additional requirements or special optimisations, and lists keeping Business Profile information up to date among good practices. For ChatGPT, OpenAI asks that sites do not block OAI-SearchBot. The rest comes from testing.
- A complete Google listing per location: right category, hours, services, photos of the branch rather than the brand. It is the most cited source in both our test and BrightLocal's study.
- One website page per location, with the listing's address, that branch's services and a reason to choose it. Not ten identical pages with the town name swapped.
- Reviews above your category's threshold: in our test, 4.8 stars for dentists and accountants, with replies on each branch's listing.
- The two or three directories AI cites for your sector, with the same details as the listing. You find them by asking the questions and reading the links, as we did.
- Structured data that matches the visible text, one per location, and OAI-SearchBot free to read the site.
- Measure by town, every month, with the same questions: a single answer is an example, not a data point.
Steps 1, 2 and 4 are the ones that change citations. A local answer leans on sources that state the address, hours, rating and services of a specific branch: if that information exists only for head office, the other branches give AI less to work with. More consistent pages and listings mean more chances to be named or linked, with no guarantee, because answers change from one run to the next. Why a page becomes the link under the answer is covered in AI brand mentions; if you want us to do it branch by branch, see our AI monitoring service, with the content side in GEO.
For a monthly write-up with one row per location, there is a template in our AI visibility report guide. To compare it with Maps rankings you also need local SEO tools.
Frequently asked questions
It depends: OpenAI says it relies on third-party business listing providers for local businesses, without naming them all. In our test on 27 October 2026, 5 answers out of 9 linked to Google Maps and 8 out of 9 showed listings with ratings and review counts.
No. Not necessarily: in our test only 6 of the 32 businesses ChatGPT named were also in the local pack, and for dentists in Turin none of the four were.
Yes. In our test they appeared on all 9 searches and sat above the local pack on two. On the Milan accountant search, 20 of the 24 references were Google Business Profiles.
Yes. In our test chains got in one branch at a time, through that branch's listing and page; a single page with ten addresses gives AI very little text about each one.
It depends: on the number of services and on competition in each town. A starting point is 10 questions per location, always worded the same way with only the town changing: 50 questions a month, repeated monthly.
Summary
9 searches, 3 cities, 32 businesses named by ChatGPT.
Only 6 were also in the local pack. 20 were still on Google's first page.
AI Overviews on 9 searches out of 9, built mostly on Google listings.
Every recommended dentist and accountant had 4.8 stars or more.
Chains get in one branch at a time, through that branch's listing and page.
There is a side effect few people consider. For years a "find us" page with ten addresses was enough, because Google Maps did the work of matching the search to the right branch. An AI answer makes that match by reading text: if your York branch has no page of its own, to ChatGPT it is just an address in a list, and an address in a list rarely gets recommended.
