Studies / Toronto
We asked ChatGPT, Gemini and Google's map for dentists in 186 areas of Toronto, more than 14,000 times. Every neighbourhood got its own answer, and neighbours next door can share no dentists at all.
What we found
To AI, every neighbourhood is its own market.
158 → 158
We asked about all 158 official neighbourhoods. On most platforms, none merged with another.
Finding 01 ↓Next door is a different market.
0
Bloor West Village and Etobicoke sit side by side, yet the AIs named no dentist for both.
Finding 02 ↓Who gets named changes within weeks.
~50%
About half the dentists ChatGPT and Gemini named were different six weeks later.
Finding 06 ↓ChatGPT, Gemini and Google's map results. All 158 of the City of Toronto's official neighbourhoods, plus the names locals actually use (the Annex, King West, the Danforth, Liberty Village and others).
Why it matters
If AI answered every neighbourhood with the same city-wide list, one measurement would cover the whole city. It doesn't. A practice can be named again and again for its own neighbourhood and not at all two neighbourhoods over. A clinic that tops the answer for "Toronto" can be missing from the answer for the street its patients actually live on.
Most businesses that serve a big city haven't thought about this, and it matters most to the ones with the most at stake:
Groups with several locations
Each location lives or dies in its own neighbourhood's answer, not the city's.
Businesses opening or moving
The neighbourhood you pick decides which answer you're competing to be in.
Agencies and marketers
A client's city-wide number can look fine while their own streets tell a different story.
It isn't only dentists. Any business people look for "near" somewhere, in a city big enough to have neighbourhoods, has the same question to answer.
What the study delivered
In a big city, you first have to know which areas to ask about, and where one market ends and the next begins. This study built that map: for every area we asked about, which practices the AIs name, how steady those answers are, and whether the area behaves as a market of its own.
Who's named, area by area
The practices each platform names for each of 186 areas.
Where markets begin and end
Which areas share answers and which stand alone.
Who shows up across town
The practices named in many neighbourhoods, not just their own.
It's the groundwork for showing any practice, group or chain who AI names in each neighbourhood it serves.
Along the way
Patterns from the map. No practice is named.
01
We expected Toronto's 158 official neighbourhoods to collapse into a few dozen bigger markets. They didn't. Ask about 158 neighbourhoods and, on most platforms, no two of them merge into one market. The only areas that merged were two names for the same place, like "the Annex" and "Annex".
158 → 158
neighbourhoods asked about → distinct answers
02
Bloor West Village and Etobicoke sit side by side, and the AIs didn't name a single dentist in common for them, on any platform. King West and Liberty Village are a short walk apart and shared almost none. Even downtown, the Fashion District, the Entertainment District and "Downtown Toronto" each got their own list.
Bloor West Village
vs Etobicoke
0
dentists in common
03
Across our asks, a neighbourhood with two dentists inside its borders still gets a dozen or so dentists named, drawn from the streets around it. A neighbourhood with fifty gets about the same. The size of the answer barely tracks the size of the market.
~11
dentists ChatGPT named per area across 25 asks, on average, whether the area has 2 or 50+
04
Simple names that mean the same place got the same dentists: "the Junction" and "Junction Area", "Roncey" and "Roncesvalles". But the city's longer official names didn't line up with the name people actually use. "Palmerston-Little Italy" and "Little Italy", or "Kensington-Chinatown" and "Kensington", returned noticeably different lists.
05
For three neighbourhoods, Google showed no map results at all, in all 25 tries each. For two more, it mostly showed none. ChatGPT and Gemini named dentists for every one of the 186 areas, every single time.
06
We asked about seven of these areas in August and again six weeks later. On ChatGPT and Gemini, roughly half the dentists named had been replaced. Google's map changed less, about a quarter. Within each round the answers were steady; between rounds, they moved.
~½
of names replaced in six weeks on the chat AIs
07
Of the 79 clinics belonging to Canada's two largest dental groups, 66 were named at least once by ChatGPT or Gemini. Of the 1,177 independent practices with their own website, 730 were. We didn't adjust for size, age, reviews or marketing spend, so this shows who was named, not why.
84%
of chain clinics named (66 of 79)
62%
of independents with a website named (730 of 1,177)
Which practices were named in which neighbourhoods, and how often, isn't published. Each practice or group can get its own private results.
We can measure how AI answers for every area you serve, in Toronto or any other city.
This study used ChatGPT, Gemini and Google's map results; Claude and Perplexity weren't part of it. The final analysis was completed on 26 October 2026. ChatGPT changed its answer format partway through the study; we recorded it and checked that it doesn't change the headline. The six-week comparison can't separate time from model changes, so we report how much the answers moved, not why. Finding 07 was recounted on 2 November 2026: "named" means named at least once by ChatGPT or Gemini, with no minimum number of asks. One group's clinics were partly identified from AI answers, which flatters its figure; the other group on its own comes out at 81% (38 of 47).