Why ChatGPT Leans on Google Maps for Local Answers

Here is the single number this entire page rests on. Across nine Ontario cities and five service verticals, ChatGPT cited google.com 130 times. The next-most-cited domain in its answers, a Burlington personal injury firm, appeared exactly twice. After that, a long tail of individual business sites, each cited once. One domain did more than 60 times the work of its nearest rival. No other source comes close, and the gap is not subtle. It is the most lopsided result in our whole dataset, and it rewrites the advice most Ontario owners have been given about ChatGPT.

That advice, almost everywhere, is some version of "ChatGPT uses Bing" or "Foursquare powers most local answers." Local Falcon lists Bing Places as the number-one source. Other widely shared posts claim Foursquare supplies over 70 percent of local results. BrightLocal's careful 2024 study went the other way and reported that Google Maps does not even appear as a directory source. These claims cannot all be right, and our data says the prevailing ones are now wrong. We have the raw transcripts to show it.

ChatGPT Runs a Live Web Search Before It Answers a Local Prompt

ChatGPT does not answer "best HVAC company in Burlington" from memory. It searches the live web first, then writes the answer from what it retrieves. This is the foundation everything else sits on, and it is more settled than it was a year ago. Ricardo Batista's cloro grounding study, published in January 2026, ran 100 prompts from independent residential-IP sessions across 52 countries and measured live web search in 80.5 percent of successful ChatGPT responses, far above the 20 to 40 percent that earlier commentary assumed.

For local queries the share is effectively total, because a model cannot know today's opening hours or this week's review count from a training snapshot. It has to look. That single fact reframes the optimization problem. You are not trying to get your website memorised by a model. You are trying to win the live retrieval step, and the question that matters is simple: when ChatGPT looks, what does it actually look at?

That Live Search Resolves to google.com, 130 Citations to 2

When ChatGPT looks for a local business, it looks at Google. We ran our scrape through DataForSEO's LLM endpoints against the live engine in May 2026, asking the same shape of question, "who are the best {vertical} in {city}, Ontario," across Brantford-area and Greater Toronto cities including Burlington, Guelph, Cambridge, Hamilton, Kitchener, Mississauga, Oakville, Toronto, and Waterloo. We logged every domain ChatGPT cited. The frequency table is not close.

Domain ChatGPT cited	Times cited	Cities	Verticals
google.com	130	7	5
bergellaw.com	2	2	1
tenacityhvac.ca	1	1	1
wmgreenroofing.ca	1	1	1
westwooddentalgroup.ca	1	1	1
ultracomfortguelph.ca	1	1	1

Read the spread, not just the top row. google.com is the only domain ChatGPT reached for across seven of the nine cities and all five verticals. Everything else is a single business website that happened to get named once, in one city, in one trade. google.com is structural; the rest is noise. When ChatGPT grounds a local answer, it grounds it in Google's local layer, which means Google Maps and the Google Knowledge Graph. That is not a directory in the sense BrightLocal meant in 2024. It is the index Google itself maintains of every claimed business, and ChatGPT is reading it back to you.

This is independently corroborated outside our dataset. Local Dominator's April 2026 report looked at 267,280 AI citations across six engines, ChatGPT, Gemini, Perplexity, Claude, Grok, and Google AI Mode, and found google.com the single most-cited platform overall with 55,977 citations. Our Ontario-only ChatGPT slice and their cross-engine national set point at the same conclusion from different angles: for local answers, Google is the ground truth the machines read.

The reason we lead with this number rather than burying it is that AI engines themselves reward front-loaded evidence. Kevin Indig's early-2026 Growth Memo analysis of ChatGPT citations found that about 44 percent of AI citations come from the first 30 percent of a page. His broader work establishes that ChatGPT cites real, crawlable sources rather than inventing them. Our scrape extends that into local search and names the source: for "best {trade} in {city}, Ontario," the crawlable source ChatGPT keeps returning to is google.com.

The Raw ChatGPT Answer Renders Your Google Business Profile Card

The clearest proof that ChatGPT is reading Google is that the answer looks like Google. When we captured the full text ChatGPT returned for "best HVAC company in Burlington, Ontario," it did not paraphrase a website. It rendered the fields of a Google Business Profile, one after another, for each business. Here is the real transcript, lightly trimmed, for the top result it named.

Look at what those fields are. "Open now" is the live hours from a Google Business Profile. "4.8 (100 reviews)" is the Google rating and review count. "HVAC contractor" is the Google primary category. The street address is the verified address pin. None of that comes from the ABW Air Systems website. All of it comes from the profile ABW maintains on Google. ChatGPT is, quite literally, reading the business its profile, and the businesses it named, ABW Air Systems and Tenacity HVAC, are real Burlington companies whose Google presence put them at the front of the line.

If you have spent any time wondering whether AI search is a separate discipline from the Google work you already know, this transcript is the answer for ChatGPT. The two are the same surface. We expand on that overlap in our guide to making your Google Business Profile legible to AI engines, which is the practical companion to this analysis.

Not Foursquare, Not Bing Places: the Ontario Data Settles a Long Argument

The popular claim that ChatGPT runs on Foursquare or Bing Places does not survive contact with a real Ontario transcript. We went looking for foursquare.com and bingplaces.com in our ChatGPT citation set. Across 1,732 citations and nine cities, neither was the grounding source for these local answers. The domain doing the work was google.com, by a margin of 130 to 2. Whatever was true of an earlier ChatGPT, or whatever holds for a different prompt type, the live local behaviour we measured in May 2026 is Google-grounded.

It is worth naming why the old consensus formed. ChatGPT Search was built on a Bing-derived web index, so it was reasonable to assume the local data rode along with it. BrightLocal's November 2024 study of 800 manual searches found business websites at 58 percent of ChatGPT sources and concluded that Yelp, Facebook, and Google Maps did not appear as directory sources at all. That was a careful result, honestly reported, and it was probably true at the time. What our data shows is how fast and how far ChatGPT's local grounding has shifted toward Google in the eighteen months since. The lesson is not that BrightLocal was sloppy. The lesson is that this behaviour moves, and stale advice gets expensive.

This also corrects a sister myth in the other direction. Because Gemini and ChatGPT both touch Google infrastructure, people assume they ground identically. They do not. Our wider scrape found each engine reading a different slice of the web, a pattern we document in our cross-engine read on how each assistant grounds its local shortlist differently. ChatGPT's Google dependence is specific to ChatGPT. Do not generalise it to Claude or Perplexity, which reach for curated directories and review aggregators instead.

Relevance, Distance, and Prominence Decide the Shortlist ChatGPT Inherits

Because ChatGPT reads Google's local layer, it inherits Google's local ranking rules, full stop. Google states plainly that local results are "mainly based on relevance, distance, and prominence." Those three words are not marketing. They are the published mechanics of which businesses Google ranks, and therefore the mechanics of which businesses ChatGPT sees first when it grounds an answer.

Relevance is how well a profile matches the query, which is why your primary category and service list matter. Distance is how close the business is to the searcher's implied location, which is why the city in the prompt does real work. Prominence is, in Google's own framing, how well-known a business is, shaped by review count, review score, and the broader web presence Google can see. A business that wins on these three inside Google is the business ChatGPT pulls into its three-to-five-name answer. You are not optimizing for a new ChatGPT algorithm. You are optimizing for relevance, distance, and prominence, the same as you always were, except the payoff now lands in two places at once.

Why ChatGPT Names Only Three to Five Businesses When Google Shows Dozens

ChatGPT names a handful of businesses because a generative answer is a summary, not a results page. Google Maps can show you twenty pins and let you scroll. ChatGPT has to commit to a short, readable list, so it takes the top of Google's ranked set and writes it up in prose. In our Burlington HVAC sample it named five, led by ABW Air Systems and Tenacity HVAC. That compression is the whole game for a local business. On Google you can rank tenth and still be found by a determined scroller. Inside ChatGPT, tenth is invisible. There is no page two of a chat answer.

This is why the academic work on generative engines matters here. The peer-reviewed GEO paper by Pranjal Aggarwal, Vishvak Murahari, and co-authors (arXiv:2311.09735, KDD 2024) showed that what surfaces a source inside a generative answer is content-level and citation-level signal, and that targeted optimization can lift a source's visibility by up to 40 percent. For a local business, the source ChatGPT is grounding in is your Google Business Profile, so the visibility lever is the strength of that profile relative to the others Google ranks. Move from the eighth-strongest profile in your city to the third-strongest, and you move from invisible to named.

The Google Business Profile Fixes That Actually Move ChatGPT Visibility

The work that gets an Ontario business cited by ChatGPT is Google Business Profile work, prioritised by what Google ranks on. Because the grounding source is now unambiguous, the to-do list is concrete rather than speculative. Here is where we tell Ontario owners to spend the quarter.

This maps cleanly onto two of Formative Digital's 12 Vectors. Vector 10, Localize, is about making your local entity unambiguous to every system that reads it, and for ChatGPT that system is Google. Vector 5, Cite, is about earning placement in the sources an engine trusts, which for ChatGPT collapses to a single dominant source. We run these through the Formative Forces, our orchestrated multi-agent system, so the profile, the reviews posture, and the supporting on-site structured data are worked in parallel rather than one fix at a time.

One honest caveat, because GEO work is money work and deserves a straight answer. Outcomes depend on your industry, your competition, and how strong your Google presence already is. There is no switch that flips the same way for every business. Our Brantford retail client Mattress Miracle grew from roughly 1,000 to over 82,400 monthly organic visits (SEMrush, April 2026) through sustained structured-content work, and the owner, Brad, put it this way: "In 40 years of advertising I've never seen anything like this. It's a completely new business." That reflects one industry and one starting point. Yours will differ, which is why we diagnose a profile before we promise anything about it. The mechanism, though, does not differ: for ChatGPT and local search, the Google Business Profile is the surface, and the surface is fixable.