Reading time: 14 minutes
To get ChatGPT, Gemini and Google to recommend a local business, you need complete and consistent data wherever AI looks. That includes the Google Business Profile, a website with a page for each location, reviews and directories. However, competition among local businesses is tough: in a study of U.S. chains, ChatGPT recommended only 1.2% of locations.
In short
- AI recommends few local businesses. In a SOCi index of about 350,000 chain locations in the U.S., ChatGPT recommended 1.2%. In Google’s 3-pack, in contrast, presence was 35.9%.
- Google says Business Profile helps a local business appear in AI answers. The Gemini app also uses public Google Maps data to find businesses.
- OpenAI, however, does not name ChatGPT’s local data providers, and its documentation offers no business profile to claim.
- A complete profile, reviews and a page for each location have official backing. Consistent data in directories and Bing Places is good practice.
Why does AI recommend so few local businesses?
AI delivers a short list, not a page of results. Take a request like “melhor pizzaria em Moema” (“best pizza place in Moema”, a São Paulo neighborhood). Here, the assistant names a few options, and local businesses left out do not appear. The main study on the topic shows a much stricter selection than Google’s. Note that it comes from the U.S. and from an industry vendor.
SOCi measured this in its Local Visibility Index 2026, with 2,751 chain brands and about 350,000 locations in the U.S. SOCi also sells local marketing and AI visibility tools to chains. For each location, it asked each AI for 10 local businesses of the same type in the area. A location counted as recommended if it appeared among the first 5.
The 3-pack is the block of three local businesses on the map in Google Search. The table compares brands’ presence in it with each AI’s recommendation rate.
| Where | Presence or recommendation |
|---|---|
| Google 3-pack | 35.9% |
| Gemini | 11.0% |
| Perplexity | 7.4% |
| ChatGPT | 1.2% |
Moreover, being strong on Google does not guarantee local businesses a place in AI answers. In retail, for example, only 45% of the 20 most visible brands in local search were among the 20 most recommended by AI. Monica Ho, SOCi’s chief marketing officer, summed up the shift at the study’s launch:
“Consumers aren’t scrolling through options anymore. They’re asking AI to decide for them…”
Monica Ho, chief marketing officer at SOCi (2026)
The habit is growing fast. In the annual BrightLocal survey of 1,002 U.S. adults, 45% used ChatGPT or another AI to ask for local business recommendations. A year earlier, it was 6%. BrightLocal also sells local SEO software. But we found no equivalent number for Brazil. If the topic is new to you, start with the guide on what GEO is.
How does Google choose local businesses for AI answers?
Google does not detail how it picks local businesses for AI answers. It does document, though, that Business Profile helps a local business appear in those AI answers. In local search, ranking follows relevance, distance and prominence. So a complete profile, a consistent website and reviews are the safest foundation.
According to Google’s guide to generative AI features, Business Profiles help products and services appear in AI answers. In Business Profile Help, Google explains the three local search factors. Relevance is how well the profile matches the search. Distance is how close the business is to the person searching. Prominence is how well-known the business is, measured in part by links and reviews. According to Google, a local business with complete and accurate information is more likely to show up.
Google compiles local business information from several sources, including crawled content from the official website and user contributions. Outdated data on the website, for example, can conflict with the profile. That is why the website and the profile need to say the same thing.
To understand how AI Mode cites websites, see how AI Mode and AI Overviews choose their sources. What changes in the rest of SEO, from crawling to indexing in Bing, is in the guide to SEO and AI.
Where does ChatGPT get local business data?
OpenAI does not say who provides this data. According to the OpenAI Help Center, ChatGPT estimates location from the IP address. It may then share it with third-party providers of local information, such as nearby business listings. The documentation also offers no business profile that an owner can claim in ChatGPT.
For reservations, the same page mentions availability from OpenTable, Resy or Yelp, depending on the restaurant. OpenTable works in all countries, while Resy works only in the U.S. and Yelp in the U.S. and Canada. The guide on how ChatGPT chooses the sites it cites explains how it rewrites “near me” requests.
So, with no declared provider, the way forward is to be accurate in every likely source. Yext analyzed 6.8 million citations from ChatGPT, Gemini and Perplexity in local searches in July and August 2025. Business websites generated 44% of citations, listings 42%, and reviews and social media 8%. In OpenAI’s answers, however, listings accounted for 48.7%. Yext also sells listings management, and it used client and prospect locations from several countries, with no breakdown by country.
Bing Places is good practice: no OpenAI or Microsoft document says ChatGPT uses this data. According to Microsoft, signing up is free and puts the business in Bing search and on Bing Maps. It also lets you import your Google profiles.
How do Gemini and Google Maps use Business Profile?
The Gemini app uses public Google Maps data to find local businesses, and Maps brings together Business Profiles and reviews. According to Gemini Apps Help, it shows the address, website, ratings and opening hours of places. SOCi also found Maps to be Gemini’s main source. In Maps itself, Gemini now powers a conversational experience, Ask Maps.
Ask Maps (Pergunte ao Maps in Portuguese) launched in March 2026 in the U.S. and India. According to Google, it analyzes more than 300 million places and reviews from more than 500 million contributors. It also personalizes answers based on the places a person has searched for or saved.
At Google for Brasil on June 10, 2026, the company announced the feature in Portuguese. It would first reach Local Guides and then everyone in July. The example given for Brazil is asking for nearby places with good pastéis (fried pastries) that accept meal vouchers. On August 6, Google’s blog in Brazil wrote that the feature is rolling out in the country. Still, we found no official announcement that it had reached everyone. Either way, to appear in these AI answers, a local business needs a Google profile that says what it offers and accepts. If Google still shows an old address after a move, see how to remove outdated content from Google.
Which AI is most accurate with local business data?
Accuracy varies from engine to engine. According to SOCi, ChatGPT and Perplexity got about 68% of business contact details right. Gemini, with Google Maps as its main source, was more accurate, but SOCi does not detail how it measured this. In the Yext study, 52.1% of Gemini’s citations came from business websites. For the app’s other uses, see how to show up in Gemini.
| Engine | Source of local data | What the business controls |
|---|---|---|
| AI Mode and AI Overviews | Google’s index and Business Profile | Profile, website and reviews |
| Gemini and Ask Maps | Google Maps | Profile, photos and reviews |
| ChatGPT | Search engines and unnamed providers | Website, listings and reviews |
Do reviews influence AI recommendations?
Probably, but the evidence is indirect. Google says more reviews and positive ratings help a local business rank better, and Gemini shows reviews from Maps. In the SOCi study, for example, locations recommended by ChatGPT had an average rating of 4.3 stars. Still, the study shows association, not causation: a high rating may go hand in hand with other strong signals.
In the same study, the average business had a 4.2 rating on Google. Locations recommended by Perplexity had 4.2 stars, and those recommended by Gemini had 3.9, below that average. So SOCi reads the rating as an entry filter, not a ranking criterion. Another figure from the study: the average business responds to 46.9% of its Google reviews.
Google recommends replying, because a reply shows the business values customer feedback. In practice, then, ask every real customer for a review, reply to all of them and track the rating by location. Never buy reviews or filter who can leave one.
In Brazil, the reputation of local businesses also runs through Reclame Aqui, a consumer complaints site. The same goes for industry platforms such as Doctoralia, iFood and Tripadvisor. Even so, we found no study that measures their weight in AI local recommendations. In the Yext study, listings accounted for 52.6% of citations in healthcare, led by U.S. directories such as WebMD and Vitals. It pays to map the reviews, Reclame Aqui and other sources AI consults for the client’s industry.
Is it worth creating a page for each city or neighborhood?
It is worth creating a page for each real location, with that location’s address, phone number, hours and services. On the other hand, it is not worth creating near-identical pages for each city just to capture searches. After all, Google may treat them as doorway abuse, a type of spam.
The Business Profile guidelines call for one profile per location and a phone number or website specific to each location. As a result, the location page is the natural destination for the profile’s link. Google also says it can detect a business’s category from its website and from mentions of the business on the web.
As an example of doorways, Google’s spam policy lists multiple pages targeted at cities that funnel users to one page. In Business Profile, service-area businesses set a service area and hide their own address. On the website, then, that area can go on the service page, without multiplying URLs.
A page of your own carries weight in citations. In the Yext study, for example, websites and local pages accounted for nearly 60% of citations in objective, unbranded questions. LocalBusiness schema helps Google read the hours and address of a local business. But according to Google itself, generative AI does not require structured data.
What changes for chains and franchises?
Chains have a scale problem: each location needs its own profile, page and reviews, and they all need to say the same thing. According to SOCi, early patterns suggest that models favor single-location businesses, whose signals are simpler. For chains, therefore, consistency across locations becomes the first task.
SOCi also says that a few locations with wrong or inactive data are enough for AI to see risk in the whole brand. However, the company publishes no figures to back these readings. On Google’s side, the guidelines call for the same name across all locations in a country, with exceptions such as sub-brands. They also call for the same category for the same service. A workflow for agencies that serve chains:
- Take inventory. List every location with name, address, phone number, hours, category and URL.
- Create a single source. Profiles, website and directories should all come from the same spreadsheet or system.
- Standardize the profiles. One verified profile per location, with the same name and the same category.
- Publish one page per location with the same data as the profile and what only that location offers.
- Replicate in directories. Bing Places, Apple Business (formerly Apple Business Connect) and the directories for the client’s industry.
- Set review goals and response goals for each location, not just for the brand.
- Measure by city in the markets where the chain has locations.
Checklist: what has evidence and what is good practice
For a local business, the best-supported actions are the ones Google documents: a complete, verified profile, reviews and a page for each location. Consistent data in directories and Bing Places has support from vendor studies or is good practice. For that reason, the table separates the two for the client’s plan.
| Action | Basis | Strength |
|---|---|---|
| Complete, verified profile | Google Help and guide | Official |
| More reviews and replies | Google and SOCi | Official and correlation |
| Page for each real location | Google guidelines and Yext | Official and study |
| Same name, address and phone number | Google, SOCi and Yext | Indirect evidence |
| Bing Places | Microsoft | Good practice |
| Apple, Reclame Aqui and industry sites | No study in Brazil | Good practice |
How do you test whether AI recommends a local business?
Test with a fixed list of local requests, by city and neighborhood, repeated every month. Write them the way customers ask, such as “melhor dentista em Pinheiros” (“best dentist in Pinheiros”). Then run them in ChatGPT, Gemini and AI Mode. Record whether the local business made the list and which sources AI cited.
You can copy the SOCi template: “Can you recommend businesses of X type in Y market?”. After that, count whether the location appears among the first 5, as in the study. For chains, run the request in every city with a location.
Control for location, because ChatGPT uses the IP address and Ask Maps personalizes results by history. So use a temporary ChatGPT chat and a Maps account with no history, and note where you ran each test. In Gemini, Maps data also requires the Keep Activity setting to be on. The repetition and margin-of-error rules in the measurement guide apply to a monthly test with prompts by city. With dozens of locations, see the GEO tools and the collection country of each one. Ask the vendor whether you can set a fixed city.
What we found in 10 Portuguese-language guides on AI and local business
We analyzed 10 Portuguese-language pages on how to show up in ChatGPT and Google with a local business. None cites a study that measures how many local businesses AI recommends. Also, none mentions Ask Maps, announced for Brazil in June.
Only 2 mention Bing Places, 2 Apple Maps and 1 Reclame Aqui. Three cover multi-location chains, but only one focuses on the topic. Four recommend pages by location or area. In contrast, only one warns against duplicate local pages, and none cites Google’s doorway policy. One confuses GEO with geotargeted SEO.
Frequently asked questions
No. OpenAI’s documentation offers no business profile that an owner can claim in ChatGPT. OpenAI also does not name its local data providers. So keep accurate data on the website, in listings and in reviews, which are the likely sources.
Google announced Ask Maps in Portuguese on June 10, 2026, first for Local Guides and for everyone in July. On August 6, Google’s blog in Brazil wrote that the feature was rolling out in the country. However, we found no official announcement of a general release.
No. LocalBusiness schema helps Google read hours and addresses. But according to Google itself, generative AI does not require structured data. A complete, verified Business Profile, reviews and a dedicated page for each real location carry more weight.
No. In Business Profile, service-area businesses set a service area and hide their own address. So on the website, that area can go on the service page, without multiplying URLs. After all, Google may treat near-identical city pages as doorways, a type of spam.
Want to know how your brand shows up in AI answers today?
NotHuman is opening spots for agencies and brands to track how they appear in AI answers. Join the waitlist and we’ll let you know when a spot opens up for you.
Main sources
- SOCi, Local Visibility Benchmarks for Multi-Location Brands, with data from the Local Visibility Index (2026).
- SOCi, Local Visibility Index launch announcement (2026).
- BrightLocal, Local Consumer Review Survey (2026).
- Yext, study of 6.8 million AI citations in local searches (2025).
- Google Search Central, Google’s Guide to Optimizing for Generative AI Features on Google Search (2026).
- OpenAI, Searching the web with ChatGPT (2026).
