What original research across Canadian clinics revealed about how AI search is changing local business discovery.
Twenty-three years ago, I wrote my Master’s thesis on a decision-support system for rent-control disputes in Egypt. The project stayed at the university, but the question behind it stayed with me: how do you turn overwhelming and sometimes conflicting information into something people can actually use to make a decision?
I spent the next two decades building enterprise systems in airlines, real estate, healthcare and telecom. I always wanted to build my own company, but software required a team and budget I did not have. Then the tooling changed.
Last year, for the first time, I could design, build and ship a real product largely on my own. I became a solo founder in Milton, Ontario, and AI is a big part of the reason my company exists.
Then I started measuring what AI was doing to other Canadian businesses.
The Change I Noticed Before I Had Any Data
I saw it first in my own behaviour. I used to research a purchase by asking around, visiting websites and comparing a few options. Now I often ask an AI assistant to do the first round of research for me. My family does the same.
Turn that behaviour around and look at it from the business owner’s side. When someone asks ChatGPT, Gemini or another AI assistant to recommend a local business, the assistant may shortlist only a handful. Some businesses get named. Many others never enter the conversation.
I looked for Canadian data showing how often this was happening. I found plenty of discussion about AI search, but very little Canadian measurement.
So I decided to measure it.
We Asked AI to Recommend Canadian Clinics 2,030 Times
Our first study focused on rehabilitation clinics across eight Canadian cities and ten common conditions. We put 2,030 questions to ChatGPT, Gemini and Google’s AI Overview and recorded which clinics appeared.
The result surprised me: 69% of the 2,129 clinics in our measured set never appeared once.
Across the city-scoped results, AI produced 9,375 clinic appearances, but those appearances came from only 663 unique clinics. Demand was growing too. For those ten rehabilitation conditions, our data estimated about 47,000 AI searches per month in Canada, up about 73% year over year.
That raised another question: is AI simply reproducing Google’s rankings, or is something different happening?
So I Tested Optometry
Our next research looked at optometry clinics across the same eight Canadian cities. Before running it, I expected considerable overlap between Google’s top results and the clinics AI recommended.
Instead, we found that Google rankings and AI recommendations could behave very differently.
We asked the same 24 questions three times over two days, using identical wording. Google’s top organic result stayed the same in 22 of 24 questions. ChatGPT’s top recommendation stayed the same in 7. Gemini’s in 6. Perplexity’s in 5.
Nothing about the clinics had changed during those two days. The businesses, websites and reviews were the same. The AI answers changed anyway.
That means the familiar idea of being a stable “number one” becomes less clear in AI recommendations. Different engines can disagree, and the same engine can produce different rankings across repeated runs.
The More Useful Discovery Was Hiding in Specific Questions
The optometry research revealed something even more practical.
When we asked ChatGPT a broad question such as “Who is the best optometrist in this city?”, it used a clinic’s own website as a source for only 2.8% of the top-three recommendations we measured. When we asked specific service questions, such as dry-eye treatment or children’s myopia control, clinic websites were used as sources 43% of the time.
Both figures are from ChatGPT only, measured across its top three recommendations. Each assistant exposes and reports sources differently, so this measure should not be compared directly across engines.
The same clinics were being evaluated. What changed was the question.
That suggests a broad “best in the city” page may not always be the most useful place to focus. A page that clearly answers a specific customer need may give an AI system more useful information to work with.
There was another striking contrast. The exact phrase “optometrist near me” draws about 74,000 Google searches a month in Canada. In our AI-demand data, that exact phrase measured zero.
That does not mean people never ask AI for local optometrists. It means they do not necessarily use the same language with an AI assistant that they type into Google. The search behaviour itself is changing.
What I Would Tell Another Canadian Business Owner
You do not need to become an AI expert. Start by asking the assistants the questions your customers would actually ask. Do not test only your company name or “best business in my city.” Ask about the specific problems, services and situations that bring customers through your door.
Run the questions more than once. Look at who gets recommended and, where sources are available, what information the assistant relied on.
Then look at your own website differently. Does it clearly answer those specific questions, or does it mostly tell visitors that you provide excellent service?
Our research so far covers rehabilitation and optometry, so I would not assume every Canadian industry behaves exactly the same way. But one thing is clear: small businesses are no longer competing only for positions on a traditional search-results page. AI systems are increasingly researching, comparing and shortlisting businesses on the customer’s behalf.
I built LeapOne because AI made it possible for one person to build something that once would have required a team.
Now I am using it to understand the other side of that change:
When AI helps Canadians decide where to go, which local businesses make it into the answer?
Marwa Saleh is the founder of LeapOne, a Canadian AI-visibility platform for local small businesses, and lives in Milton, Ontario. LeapOne’s Canadian AI visibility research is published at leapone.ca.
About Author
Marwa Saleh is the founder of LeapOne, a Canadian AI-visibility platform for local businesses. She has more than two decades of experience building enterprise systems across healthcare, airlines, real estate and telecom. Based in Milton, Ontario, she now researches how AI search is changing the way Canadian consumers discover local businesses, using original Canadian data from LeapOne’s studies of rehabilitation and optometry clinics.
