Everline asked ChatGPT and Google AI Mode the question a customer would ask: who are the best businesses of this kind in this city. Then it checked which of the businesses Google ranks highly were in the answers. This page sets out the method first, then the findings, then the data.
Method
What was asked. For each of 10 local categories, one question per city, asked once of ChatGPT and once of Google AI Mode, from a logged-out browser. Each answer named five businesses. In all, 29 questions produced 58 answers and 290 business names.
| Category | Cities asked about | Listings checked, Toronto | Listings checked, other GTA |
|---|---|---|---|
| Roofing | Toronto, Mississauga, Oakville | 10 | 10 |
| HVAC and plumbing | Toronto, Brampton, Mississauga | 10 | 10 |
| Kitchen and bath renovation | Toronto, Mississauga, Vaughan | 9 | 10 |
| Dentists | Toronto, Brampton, Mississauga | 10 | 10 |
| Physiotherapy and chiropractic | Toronto, Mississauga, Oakville | 10 | 10 |
| Med spas | Toronto, Mississauga, Oakville | 10 | 10 |
| Personal injury and immigration lawyers | Toronto, Brampton and Mississauga (immigration), Mississauga and Vaughan | 8 | 11 |
| Real estate agents | Toronto, Oakville | 10 | 10 |
| Movers | Toronto, Brampton, Mississauga | 10 | 10 |
| Auto repair | Toronto, Brampton, Mississauga | 10 | 10 |
| Total | 29 questions, 58 answers | 97 | 101 |
How it was run. The answers were collected on one day in October 2026 from a logged-out browser session driven by Claude, with ChatGPT showing its “Log in” button and Google its “Sign in” button, from a Toronto connection. Each question was asked once of each assistant.
What was not recorded. The exact wording of each question and the day of collection were not written down with the data, so this edition cannot quote them. Every answer asked for five businesses, and the Rec’d Check’s first question has that shape (“Who are the best [category] in [city]? Give me 5 specific names.”), but the match is not confirmed. The next edition will log the wording, the date and time, and the location with every answer, following Everline’s live test protocol.
Which businesses were checked. 198 top-ranked Google Business listings across the 10 categories: 97 with a Toronto address and the rest in the other GTA cities asked about. Their Google rating and review count were recorded when they were sampled.
The listings came from Google’s Places data: every row carries a Google place ID and Places category types, and the sample runs ten listings per category and city, with ratings from 3.1 to 5 and review counts from 6 to 23,006. The query and the ranking used to pick those ten were not recorded, so this edition cannot state how “top-ranked” was defined beyond the sample itself. The next edition will record the query.
How “named” was decided. A listing counts as named when its business name matches a name in any answer for its category, in any city asked about. Names are matched by their distinctive words, ignoring words every business in the category shares, such as “roofing”, “dental” or “Toronto”. The script that does this is published with the data, and it reproduces the classification in the data file exactly.
What to keep in mind.
- These are listings, not businesses. 13 businesses held 30 of the listings between them. The 97 Toronto listings belong to 93 businesses, and the 30 named Toronto listings to 26.
- Name matching can miss a business named differently from its listing, or match two businesses with similar names. Every match is in the listings file so it can be checked.
- AI answers change from day to day and from person to person. This is a snapshot of one month, logged out.
- The sample is the top of Google’s results, not every business. A business outside the sample can still rank well.
Findings
Fewer than a third of top-ranked Toronto businesses were named
Of the 93 top-ranked Toronto businesses, 26 were named by at least one assistant: 28%. Counted by listing, 30 of their 97 Google listings were named. Across all 198 GTA listings, 68 were named.
| Listings | Checked | ChatGPT only | AI Mode only | Both | Neither |
|---|---|---|---|---|---|
| Toronto businesses | 93 | 11 | 7 | 8 | 67 |
| Toronto listings | 97 | 13 | 7 | 10 | 67 |
| All GTA listings | 198 | 29 | 17 | 22 | 130 |
ChatGPT named more of them than Google AI Mode
ChatGPT named 19 of the Toronto businesses and Google AI Mode named 15. Only 8 were named by both. Across the GTA, counted by listing, the gap was the same way round: 51 against 39.
The two assistants rarely agreed
For the same question, the two assistants had 1.2 names in common out of five, on average. For 8 of the 29 questions they had none. A business recommended by one assistant can be invisible on the other.
| Names both engines gave | Questions |
|---|---|
| 0 of 5 | 8 |
| 1 of 5 | 11 |
| 2 of 5 | 5 |
| 3 of 5 | 5 |
| 4 of 5 | 0 |
| 5 of 5 | 0 |
Most of the businesses they recommended were not in Google’s top results
Of the 290 names the assistants gave, 85 were among the top-ranked listings sampled and 205 were not: 70.7% of the businesses recommended were outside Google’s top results as sampled. The assistants are not reading Google’s ranking back to the customer.
Reviews helped, but did not decide it
In Toronto, the named listings had a median of 510 Google reviews against 277 for the rest. Their median rating was the same.
| Listings | Median reviews, named | Median reviews, not named | Median rating, named | Median rating, not named |
|---|---|---|---|---|
| Toronto | 510 | 277 | 4.9 | 4.9 |
| All GTA | 384.5 | 180.5 | 4.9 | 4.9 |
The share named did not rise steadily with review count, and across all GTA listings none of those rated under 4.5 was named.
| Google reviews | Listings | Named by either | Share named |
|---|---|---|---|
| Under 50 | 4 | 1 | 25% |
| 50 to 99 | 16 | 5 | 31.2% |
| 100 to 249 | 23 | 5 | 21.7% |
| 250 to 499 | 17 | 3 | 17.6% |
| 500 or more | 37 | 16 | 43.2% |
One example from the roofing data: a roofer rated 4.7 with 561 reviews was never named, while a competitor rated 4.9 was named by ChatGPT. That competitor holds two Google listings, with 67 and 255 reviews, 322 in total. The reviews question looks at this in more detail.
Categories differed a lot
| Category | Toronto listings | Toronto named | GTA listings | GTA named | Names both engines gave, of 5 |
|---|---|---|---|---|---|
| Roofing | 10 | 3 | 20 | 7 | 2.3 |
| HVAC and plumbing | 10 | 5 | 20 | 10 | 1 |
| Kitchen and bath renovation | 9 | 4 | 19 | 9 | 1.7 |
| Dentists | 10 | 1 | 20 | 4 | 1 |
| Physiotherapy and chiropractic | 10 | 3 | 20 | 6 | 1.7 |
| Med spas | 10 | 1 | 20 | 4 | 1.7 |
| Personal injury and immigration lawyers | 8 | 4 | 19 | 8 | 1.7 |
| Real estate agents | 10 | 4 | 20 | 7 | 0.5 |
| Movers | 10 | 3 | 20 | 10 | 0.3 |
| Auto repair | 10 | 2 | 20 | 3 | 0.3 |
| All ten | 97 | 30 | 198 | 68 | 1.2 |
Roofing had the most agreement between the two assistants and auto repair and movers the least. In Toronto, dentists and med spas had the fewest top-ranked listings named. Each category has its own page with the full answers:
- Roofing
- HVAC and plumbing
- Kitchen and bath renovation
- Dentists
- Physiotherapy and chiropractic
- Med spas
- Personal injury and immigration lawyers
- Real estate agents
- Movers
- Auto repair
What it means for a local business
Ranking on Google is not the same as being recommended by an assistant. The test measured the first two of the seven checks in Everline’s Seven-point check: whether ChatGPT and Google AI Mode name the business. The other five are the likely reasons when they do not, and why ChatGPT doesn’t recommend your business walks through them.
The data
Both files are linked at the top of this page.
- Every answer: category, city asked about, assistant, position in the answer, the business named, and whether it was one of the sampled listings.
- Every sampled listing: category, business name, listing city, Google rating, Google review count, and whether ChatGPT and Google AI Mode named it.
The listings file names every sampled business, including the ones the assistants did not name, so every match can be checked. Contact details, addresses and Everline’s own prospecting notes are not in the published files. To cite the data: Everline Solutions, How AI Sees Toronto Businesses, October 2026.