Georgian Bay SEO asked an AI assistant to recommend a supplier in 39 Ontario towns across 10 trades, then recorded every page the assistant read before answering. Only 6.6% of those citations were the recommended business’s own website. The rest came from directory and roundup pages the business does not control, and five domains carried 67.1% of them. Being first in the Google Map Pack made a business the assistant’s first pick 28.2% of the time.
That is the short version. The rest of this post is how Georgian Bay SEO measured it and what a Collingwood or Owen Sound owner should do about it this month.
How Georgian Bay SEO measured this
Georgian Bay SEO ran 390 buyer questions past an AI assistant with web search enabled, one per town per trade, phrased the way a customer phrases it. “Who is the best plumber in Thornbury, Ontario? Name specific companies.”
The 39 towns come from the market file this agency already maintains, from Collingwood and Thornbury out through Muskoka, Wellington, Waterloo and Halton. The 10 trades are plumber, electrician, HVAC contractor, roofing contractor, general contractor, landscaper, house cleaner, painter, auto repair shop and pest control.
For every one of those 390 questions, Georgian Bay SEO also pulled the Google Maps ranking, searched from that town’s own coordinates rather than from a province-level query. That distinction matters and it is measured further down. Total: 1,170 queries collected on August 27, 2026.
This is the local companion to the Canada AI Visibility Index 2026, which measured 368 markets nationally. The index asked who gets named. This study asks what the assistant read before deciding.
Map Pack first place buys a 28.2% chance
Every trade client eventually asks the same question. I am number one on the map, so why does the assistant not say my name.
| If you hold Google Map Pack first place in your town | Share of 390 questions |
|---|---|
| The assistant names you somewhere in its answer | 66.4% |
| The assistant names you in its top three | 50.3% |
| The assistant names you first | 28.2% |
The AI top three and the Map Pack top three share 1.26 businesses out of 3.
In Barrie, across all 10 trades, the Map Pack leader was the assistant’s first pick zero times. In Collingwood the leader was named at all in 50% of trades and named first in 10%.
The national index found the Google review leader went unmentioned in 40% of markets. This study finds the Map Pack leader goes unmentioned in 33.6% of questions. Two different definitions, two different methods, and the same order of magnitude. Map Pack rank is a real asset and it is not an AI visibility number.
Your own website is 6.6% of what the assistant reads
This is the finding that should change what a Georgian Bay trade business spends money on next.
Own-site citations, by trade
| Trade | Citations that were the business’s own website |
|---|---|
| Plumber | 1.7% |
| General contractor | 2.1% |
| Electrician | 2.8% |
| Painter | 3.4% |
| Roofing contractor | 4.4% |
| Landscaper | 5.7% |
| HVAC contractor | 6.0% |
| Auto repair shop | 8.7% |
| House cleaner | 13.8% |
| Pest control | 17.8% |
Why plumbing sits at the bottom of that table
Home services have been aggregated and re-aggregated for twenty years. When an assistant looks for the best plumber in Wasaga Beach, it finds a dozen pages that have already ranked the plumbers, and it reads those instead of reading eleven plumbing websites.
The trades where your own website still earns citations
Pest control at 17.8% and house cleaning at 13.8% sit at the top because the directory layer in those categories is thinner. A pest control company that describes its treatments, its seasons and its service radius in plain sentences still gets read directly.
Homepage rewrites will not fix a plumbing result
Georgian Bay SEO will not sell a website rewrite as an AI visibility fix in the trades at the bottom of that table. The data does not support it. Fix the pages that are actually being read.
Five pages decide most Georgian Bay AI answers
82 distinct third party domains appeared across the 390 answers. Five of them carried 67.1% of the citations.
| Domain | Share of third party citations |
|---|---|
| homestars.com | 27.3% |
| reviews.birdeye.com | 13.9% |
| proscore.ca | 10.2% |
| threebestrated.ca | 9.2% |
| bbb.org | 6.5% |
HomeStars alone is more than a quarter of everything the assistant read. Most owners in Meaford or Owen Sound have never opened their own HomeStars page and cannot say what rating sits on it.
Checking those five, correcting the trade category, the service area and the phone number on each, and getting a handful of recent reviews onto the two that carry them, is an afternoon of work. On this evidence it is the highest-leverage afternoon available in the trades.
Six in ten businesses named are not in your town
58.8% of the businesses an assistant named for a town were not located in that town. The effect gets worse the smaller the town.
| Town | Named businesses located elsewhere |
|---|---|
| Thornbury | 92.7% |
| Creemore | 89.4% |
| The Blue Mountains | 88.7% |
| Stayner | 77.8% |
| Meaford | 75.0% |
| Wasaga Beach | 64.6% |
| Collingwood | 41.5% |
| Owen Sound | 31.4% |
| Barrie | 20.4% |
Read that table twice, because it works in both directions.
A Thornbury contractor loses most of its own town’s AI answers to firms in Collingwood and Owen Sound. A Collingwood contractor is already winning answers in towns it has never targeted, and does not know it.
The towns where the leader is least visible
The Blue Mountains and Thornbury both sit at 30%, meaning the Map Pack leader went unnamed in seven of ten trades. Blue Mountain scoring badly here matches what the Collingwood review index work found from a completely different method, which is the kind of agreement worth trusting.
Review count is not the AI ranking signal
The assistant named a first-choice business in each answer, and the same answer carried Google ratings for the businesses in that town.
| Measure | Median across 390 questions |
|---|---|
| Google review count of the business named first | 54 |
| Google review count of the most-reviewed business in the same answer | 230 |
The assistant is not sorting by review count. A firm with a review count of 54 beats a firm with 230 more often than not, because the assistant is reading descriptions rather than counting stars.
Reviews still do the work they always did. Reviews win the Map Pack, and reviews close the customer who lands on your profile. Georgian Bay SEO sells review work on those two outcomes, and this study is the reason it does not get sold as an AI visibility fix.
Proximity: the Map Pack has no single answer
Georgian Bay SEO pulled the Map Pack twice for every question. Once searched from the town’s own coordinates, and once as a province-level query with the town name typed into the search box.
Those two rankings named the same business in first place 45.1% of the time.
That is the proximity effect, measured on 780 paired queries rather than asserted. It also means any Map Pack position quoted to you without a stated search origin is close to meaningless. Ask where the search was run from. If the answer is “from the office”, the number describes the office.
What to do this month
Four steps, in order. The first two cost nothing but time.
Ask the question yourself. Open an AI assistant and type the question your customer types, naming your trade and your town. Read the sources it cites, not only the answer. That citation list is your real competitive set in this channel.
Count how many of those sources are your own domain. In the trades at the bottom of the table it will usually be zero. That tells you where your money should go.
Fix the five pages above. Category, service area, phone number, and a current rating on the two that carry ratings.
Then check whether the assistant names you at all. Georgian Bay SEO runs a free AI visibility check that asks the assistants about your trade and your town and sends back the screenshots: named, not named, or a competitor named instead.
What this study does not show
One assistant, one week, one country setting. The Canada AI Visibility Index covers four assistants and is the better source for cross-assistant differences. This study is a snapshot until Georgian Bay SEO re-runs it, and the scripts are built to re-run on the same roster so the next edition is comparable.
Ten trades are ten trades. The gradient from plumbing at 1.7% to pest control at 17.8% is clear across 39 towns, and it does not tell a dentist or a bookkeeper where they sit.
45.6% of the businesses the assistant named arrived with no link attached. That looked at first like a large population of trades with no website. Cross-checking each against the Google Maps record for the same business put the real figure near 3%. The missing links are the assistant’s formatting, not an absence of websites. Georgian Bay SEO reports the corrected number rather than the alarming one.
Frequently asked questions
Does ranking first in the Google Map Pack make an AI assistant recommend me?
It helps and it does not decide. Across 390 questions in 39 towns, the Map Pack leader was named somewhere in the answer 66.4% of the time, appeared in the assistant’s top three 50.3% of the time, and was named first 28.2% of the time. In Barrie the Map Pack leader was never the first pick across ten trades. Map Pack position and AI visibility are two separate scoreboards and both are worth holding.
If the assistant is not reading my website, what is it reading?
Directory and roundup pages. Only 6.6% of the citations behind these 390 answers pointed at the recommended business’s own site. Five domains carried 67.1% of the rest: HomeStars at 27.3%, Birdeye at 13.9%, ProScore at 10.2%, ThreeBestRated at 9.2% and the Better Business Bureau at 6.5%. In plumbing the own-site share drops to 1.7%.
Should I rewrite my website to get recommended by ChatGPT?
Not if you are a plumber, an electrician, a painter, a roofer or a general contractor in this market. Those five trades all sit under 5% own-site citations, so the pages deciding your result are pages you do not own. Fix those first. House cleaning and pest control are the two trades in this study where your own site earns enough citations to be worth the rewrite.
Why do businesses from other towns keep showing up in my town’s answer?
Because AI answers ignore municipal boundaries far more than the Map Pack does. 58.8% of the businesses named for a town in this study were located somewhere else, rising to 92.7% in Thornbury and 88.7% in The Blue Mountains. It costs a small-town business its home answers. It also means a Collingwood or Owen Sound business can win answers in surrounding towns without opening a second location.
How many Google reviews do I need before an AI assistant names me?
There is no threshold, and review count is not what the assistant sorts on. The median Google review count of the business named first in these answers was 54, against 230 for the most-reviewed business appearing in the same answer. Reviews win the Map Pack and close the customer who reaches your profile, which is why Georgian Bay SEO still works on them, but they are not the lever on this particular channel.
Want this run against your own business?
Free 15-minute audit. We pull up Google together and show you exactly where you rank, where you could be in 60 to 90 days, and the cheapest moves to get there.
Prefer to talk? (705) 539-0398
