Studies / Ontario downtowns

Where does AI point people in an Ontario city?

When people ask AI where to eat, shop or spend a day in a city, how much of the answer points downtown? We asked in five Ontario cities: Barrie, Cobourg, Guelph, Kingston and Stratford.

The short answer

In all five cities, AI named downtown places far more often than downtown's size would suggest. In Kingston, only about 1 in 5 of the city's places are downtown. But about 3 in 5 of the places AI named were downtown.

1,200 asks · 5 cities · 6 platforms · November 8–9, 2026 · a first look, from two sittings

What this means for a BIA

  • AI points people downtown far more than downtown’s size would suggest. It happened in all five cities, on every AI we tested. Finding 01 ↓
  • It leans downtown most for restaurants and shops. For things to do, the gap was smallest. Finding 03 ↓
  • The wording decides what Google shows. Searches for restaurants, coffee, brunch and “best shops” got Google’s map. Searches for things to do and weekend plans got an AI answer. Finding A ↓
  • The website AI used most was usually the city’s own tourism site. In most of the five cities, the downtown association’s own website was in the top ten too. Finding C ↓

If you run a downtown association, a main street program or a business district anywhere, the same questions apply to you. They also matter to the businesses inside it, to businesses elsewhere in the city that AI names less often, and to the city’s tourism and economic development offices.

What we found

Three findings about downtowns

01

In every city, AI named downtown far more than downtown's size would suggest.

We put every place the AIs named on a map and checked whether it was inside the city's official downtown area (its business improvement area, or BIA). Then we compared that with how many of the city's restaurants, cafés, bars, shops and attractions are downtown.

Green: the share of places AI named that were downtown. Grey: downtown's share of all the city's places. Five AI platforms, 200 asks per city, November 8–9, 2026.

Barrie

Places named that were downtown34% · 422 of 1,251
Downtown's share of the city's places9% · 318 of 3,361

Cobourg

Places named that were downtown63% · 768 of 1,223
Downtown's share of the city's places32% · 186 of 578

Guelph

Places named that were downtown46% · 611 of 1,330
Downtown's share of the city's places14% · 368 of 2,656

Kingston

Places named that were downtown62% · 905 of 1,463
Downtown's share of the city's places20% · 477 of 2,407

Stratford

Places named that were downtown64% · 906 of 1,405
Downtown's share of the city's places31% · 266 of 866

A place counts each time it's named. Downtown is the area of each city's BIA: the Downtown Barrie BIA, the Cobourg Downtown BIA, the Downtown Guelph Business Association, the Downtown Kingston BIA and the Downtown Stratford BIA.

02

Every AI we tested did it, in every city.

We checked each AI (ChatGPT, Claude, Gemini, Perplexity and Google's AI answer) in each city separately. Every one named downtown more than its share, in every city. Even the lowest result, 26% of places named downtown, came from a city where only 9% of places are downtown.

Each AI in each city is based on 40 questions, so single results are rough. But even at the low end of their likely range, every one is above downtown's share.

Checks where AI named downtown more than its share (5 AIs × 5 cities)

25 of 25

03

AI leans downtown most for restaurants and shops, least for things to do.

Across the five cities, downtown has 20% of the restaurants, cafés and bars, but 63% of the ones AI named were downtown. Downtown has 15% of the shops, but 59% of the shops named. For things to do, like museums, parks and theatres, it was 12% and 35%, the smallest gap of the three.

Green: the share of places AI named that were downtown. Grey: downtown's share of all such places. All five cities together, November 8–9, 2026.

Restaurants, cafés and bars

Named downtown63% · 1,959 of 3,099
Downtown's share of places20% · 706 of 3,543

Shops

Named downtown59% · 791 of 1,339
Downtown's share of places15% · 681 of 4,432

Things to do

Named downtown35% · 661 of 1,879
Downtown's share of places12% · 228 of 1,893

What this shows about AI

Five things we learned about the AIs themselves

Asking the same questions in five cities also shows how each platform behaves.

A

Google answers food questions with a map, and weekend plans with AI.

When we searched Google for restaurants, coffee, brunch or "best shops", it showed its map every time (all 125 searches) and an AI answer only 14 times. When we searched for "things to do", "what to do this weekend" or "where to go shopping", it never showed the map, and gave an AI answer 72 times out of 75. Small wording changes matter: "where to go shopping" got an AI answer all 25 times, while "best shops" got the map.

What Google showed for each question, out of 25 searches (five cities × five times). November 8–9, 2026.

best restaurants in [city]

Map shown25 of 25
AI answer shown2 of 25

where should I eat in [city]

Map shown25 of 25
AI answer shown2 of 25

best coffee shop in [city]

Map shown25 of 25
AI answer shown2 of 25

best place for brunch in [city]

Map shown25 of 25
AI answer shown5 of 25

best shops in [city]

Map shown25 of 25
AI answer shown3 of 25

where to go shopping in [city]

Map shown0 of 25
AI answer shown25 of 25

things to do in [city]

Map shown0 of 25
AI answer shown22 of 25

what to do in [city] this weekend

Map shown0 of 25
AI answer shown25 of 25

B

Ask a different AI, get a different-length list.

Claude named about 11 places per answer and Gemini about 10. Perplexity named about 8 and ChatGPT about 4.

Claude
10.7
Gemini
9.8
Perplexity
7.6
ChatGPT
4.0

Places named per answer, 200 answers each, November 8–9, 2026.

C

The website AI used most was usually the city's own tourism site.

In three of the five cities, the website the AIs used most as a source was the city's tourism site: visitkingston.ca (in 113 of 200 asks), experiencecobourg.ca (104) and tourismbarrie.com (86). In Stratford it was the Stratford Festival's site, stratfordfestival.ca (72), ahead of tripadvisor.ca (65) and the city's tourism site, visitstratford.ca (49). In Guelph it was a local news site, guelphtoday.com (55). In most of the five cities, the downtown association's own website was also among the ten sites used most.

Cities where the website AI used most was the city's own tourism site

3 of 5

D

AI suggests events and trails, not just places.

Asked where to go, the AIs also named 355 events, tours and trails, such as festivals and walking routes. We left those out of the downtown counts, because they don't have a single address.

Events, tours and trails the five AIs named

355

E

Google's map results lean downtown too.

When Google showed its map instead, an even bigger share of the places it listed were downtown, in all five cities. Treat this one as rough: the map lists only about three places per search.

What we asked

Eight everyday questions, asked the same way in every city:

  • best restaurants in [city], Ontario
  • where should I eat in [city], Ontario
  • best coffee shop in [city], Ontario
  • best place for brunch in [city], Ontario
  • where to go shopping in [city], Ontario
  • best shops in [city], Ontario
  • things to do in [city], Ontario
  • what to do in [city], Ontario this weekend

These are first-look wordings. A full study sets its questions from evidence of how people in that city actually ask, pilots several wordings, and publishes them.

How we asked

  • Six platforms: ChatGPT, Claude, Gemini, Perplexity, Google's AI answer and Google's map results.
  • The free versions: the models each platform gave free users at the time (ChatGPT gpt-6-luna, Claude Sonnet 5.5, Gemini 3.5 Flash-Lite, Perplexity Sonar).
  • From the city: ChatGPT, Claude and Perplexity were told the asker was in the city. Gemini was asked without a location; the city is in every question. Google searches were run from each city.
  • Five times each: every question asked five times on every platform, in two sittings on November 8, 2026: the afternoon and the evening (Eastern time). Some of the second sitting's Google searches ran the next morning, November 9.
  • Every place mapped: each place named was matched to its Google listing, or found with a Google Maps search, then tested against the BIA's boundary. Barrie, Cobourg and Kingston publish their BIA boundaries as open data. For Guelph and Stratford we traced the boundary from the BIA's own published map.
  • Checked by hand: we matched 93% of the places named to a location (6,672 of 7,184). A person checked 50 matches by hand: 49 were right, 1 was unclear, and none was wrong.
  • Downtown's share of places: every restaurant, café, bar, shop and attraction Google lists within 5–8 km of downtown, depending on the city's size, and the share of them inside the BIA.

What a study shows a downtown

The same method, for your district, at full strength

Useful for downtown associations and business districts that want to know where AI points people, and how that changes from quarter to quarter.

Your share of the answers

How much of what AI names sits inside your district, by platform and by question, accurate to about ±6 points per platform.

Which places get named

The places AI names in and around your district, with counts, and how many of your district's places are never named.

What AI reads

The websites behind the answers, and whether yours is one of them.

How it moves

Tracked each quarter, as the AIs change.

Get your own study

Names stay private

No business is named on this page. Each downtown's detailed results go only to that downtown. We don't sell placement or fixes. We measure.

New findings by email

Only when a study turns up something new.

Want to know where AI points people in your city?

The same method works for any downtown, main street or business district.

Get your own study

A note on our numbers

These are patterns we observed on November 8–9, 2026, not causes, and AI answers change. This is a first look: two sittings, 200 AI asks per city, 40 per platform. That's lighter than the standard we use for paid studies, so single-platform results are indicative. A place counts each time it's named. Places we couldn't locate (7%) are left out of the shares. Even counting every one of them as outside downtown, the downtown share stays above downtown's share of places in every city (Barrie 31%, Cobourg 57%, Guelph 43%, Kingston 57%, Stratford 61%). Google's AI answer counts only when Google showed one with text (86 of 200 searches; one more showed an empty box). The AI figures in the first three findings exclude Google's map results, which are reported separately.

How we measure →