Guide · AI Search

How AI Assistants Choose Which Local Businesses to Recommend

By Alex Luna, founder of WebSEO.pro · Updated

Short answer

AI assistants recommend local businesses by searching the web and map data, then naming the companies that show up most often in trusted places. ChatGPT and Copilot lean on web search partners, Gemini and Google AI Mode lean on Google data, and Siri leans on Apple Maps. Websites, third party mentions, reviews and licensing records all matter.

Key takeaways

  • Each assistant pulls local data from a different source, so one listing is never enough.
  • In a BrightLocal study from November 2024, business websites made up 58% of ChatGPT local sources.
  • Review volume tracks with being named: an October 2026 8ight study saw a big gap between firms with few reviews and many.
  • BBB and state licensing pages showed up often in ChatGPT answers about local trades.
  • Answers change between repeat questions, so judge visibility over many prompts, not one.

How does an AI assistant decide which contractor to recommend?

It runs one or more web searches, reads the pages that come back, and names the businesses that appear most often in sources it trusts. Think of it as a fast researcher reading the first pages it finds, not a directory with a fixed ranking.

OpenAI explains this in its ChatGPT search help article. It says ChatGPT search sometimes partners with other search providers and typically rewrites your question into one or more targeted queries that it sends to those providers. It also says ChatGPT ranks results using multiple factors and that placement is not guaranteed. The other assistants work in a similar way, but they search different indexes and map databases.

For a contractor, that means your visibility depends on two things. First, can the assistant find you in the index it uses? Second, do enough trusted pages describe you as a good fit for the job and the city in the question?

Where does each AI assistant get local business data?

Each assistant draws on a different mix of search and map data. The table below sums up what each company has said in public, and where the public record is thin.

ChatGPT is the hardest to pin down. Microsoft announced in May 2023 that Bing would become the default search engine for ChatGPT. Today OpenAI's help center says only that it partners with search providers and may share your location with trusted third party providers for local results, without naming them. Copilot is Microsoft's own assistant and runs on Bing search, so Bing Places and the Bing index matter for both.

On the Google side, Google announced in October 2025 that developers can ground Gemini API answers in Google Maps data covering more than 250 million places, including user reviews and hours. That announcement covers the API, not the consumer Gemini app, so treat it as a strong hint rather than proof. Whitespark's May 2026 guide to AI Mode reports that AI Mode summaries show star ratings, review counts and hours that come from Google Business Profiles or other sources.

Apple said in January 2023 that Apple Business Connect lets businesses customize how their information appears to more than a billion Apple users across Apple Maps, Messages, Wallet, Siri and other apps. Perplexity says its PerplexityBot crawler is designed to surface and link websites in Perplexity search results.

What sources show up most when ChatGPT answers local questions?

Business websites come first, then mentions on other sites, then directories. BrightLocal ran 800 manual local searches in ChatGPT in November 2024 and recorded the first ten sources for each.

  • Business websites made up 58% of all local sources.
  • Business mentions, such as news, blogs and best of lists, made up 27%.
  • Directories made up 15%.

Which trust signals show up in AI answers about contractors?

Better Business Bureau pages, government licensing sites and review volume all show up clearly. The best recent data comes from 8ight.AI, a company that sells AI visibility services, in a study published October 5, 2026 and reported by PPC Land.

8ight asked ChatGPT and Claude through their APIs, with web search on, "Who is the best {trade} in {city}? Name specific businesses" across 39 trades in 49 places from September 10 to 30, 2026. It compared 1,412 ChatGPT answers and 218 Claude answers with 2,976 businesses found in Google Maps.

  • ChatGPT named 3.4% of tested firms with under 25 Google reviews, and 22.2% of firms with 1,000 or more.
  • ChatGPT cited the BBB in 38.7% of answers and government or licensing sites in 25.1%.
  • ChatGPT cited Yelp in 0% of answers, while Claude cited Yelp in 92.7%.
  • The study notes that API answers may differ from what the consumer apps show, and it did not test whether profile changes move results.

Where each AI assistant gets local business data, based on public statements

AssistantMain local data sourceWhat to claim or allow
ChatGPTWeb search partners (Bing named as default in 2023) plus third party local providersBing Places, strong website, allow OAI-SearchBot
Microsoft CopilotBing search and Bing PlacesBing Places and Bing Webmaster Tools
Google GeminiGoogle Search and Google Maps dataGoogle Business Profile
Google AI ModeGoogle Search index and Business Profile detailsGoogle Business Profile, indexed service pages
PerplexityIts own web results via PerplexityBotAllow PerplexityBot in robots.txt
SiriApple Maps place cardsApple Business Connect

Why do AI assistants give different answers to the same question?

Answers are generated fresh each time, from a fresh search, so the list of names shifts. In the same 8ight study, when ChatGPT got the same question twice, a median 29% of the businesses named in either answer appeared in both. The median gap between the two asks was 39 minutes.

Small changes in wording, location, the model version and which pages the search returns can all change who gets named. That is why a single screenshot of ChatGPT naming a competitor tells you little. A fair test asks the same set of questions many times, across several assistants, and counts how often you appear.

What can a contractor control to get recommended more often?

You control your website, your listings, your reviews and how much of the web talks about you. Those are the same inputs the assistants read.

  • Build a page for each service and city you want, with your license number, service area and real project details.
  • Claim and complete Google Business Profile, Bing Places and Apple Business Connect with the same name, address and phone.
  • Keep your CSLB license and BBB profile current, since licensing and BBB pages show up in ChatGPT answers.
  • Ask every happy customer for a review, steadily, on Google first.
  • Earn mentions on local news, supplier sites, trade groups and community pages.
  • Allow OAI-SearchBot and PerplexityBot in robots.txt so those engines can read your site.

How should a San Diego contractor test AI visibility?

Write ten questions your customers really ask, such as "best roofer in Chula Vista" or "who installs artificial turf in San Diego", and run each one several times in ChatGPT, Gemini, Perplexity, Copilot and Google AI Mode. Log who is named and which sources are cited.

Repeat the same test every month. Over time you will see which sources keep appearing for your trade and city, and those are the places to earn a listing or mention. At WebSEO.pro in Chula Vista we run this kind of test for contractors, and our free audit shows you where you appear today and which gaps to close first.

Sources

Related services and trades

Questions about this guide

More guides

Want this done for your business?

Get a free visibility audit that checks Google, Maps and AI search. No contract, no pressure.