We Asked Three AI Models 90 Times Who to Hire for Web Design in Niagara. We Were Named Once.

I run a two-person web design studio in Thorold, Ontario. I wanted to know something that felt slightly embarrassing to ask: if someone in Niagara asks an AI who to hire to build a website, does it name us?

So I ran the test. The short answer is that it named us once in 90 answers.

This post is how I ran it, what the models cited instead, and what I've changed since.

How I ran it

Ten questions, written the way a business owner would type them:

1. Who should I hire to build a website for my small business in Thorold, Ontario?
2. What are the best web development companies in the Niagara Region, Ontario?
3. best web designers near Thorold, Ontario
4. affordable website design for small businesses in the Niagara Region
5. who builds websites for local businesses in St. Catharines or Niagara?
6. top rated web development agencies in Niagara
7. I need a new website for my business in Niagara Falls, who should I call?
8. local web design company near me in Ontario's Niagara Region
9. ecommerce website developers in the Niagara Region
10. where can I find a freelance web developer in Thorold, Ontario?

Three models: Claude Sonnet 5, Gemini 3.8 Flash and GPT-5.6. Each question went to each model three times. Ten questions, three models, three runs each is 90 answers.

Two things you should know about the method before you trust any number below.

First, I gave every model the same web search tool rather than using each company's own app. Every one of the 90 answers used it. That means this measures what the models find and recommend when they search, not exactly what you'd see in the ChatGPT or Gemini app.

Second, a separate model pulled the business names and cited pages out of each answer, and I haven't hand-checked every extraction. The same business can be split across a couple of spellings. Our own count is unaffected, because we were mentioned so rarely, but treat the individual business counts as close, not exact.

We were named once

The Biz Spark appeared in 1 of 90 answers. It was Gemini, on "best web designers near Thorold, Ontario." Claude never named us and neither did GPT.

It's worse than one mention suggests. Across the 175 different pages the models cited, none was on thebizspark.com.

A few names take most of the answers

The models named 84 different businesses. Forty-two of them were named once. At the other end, the most-named business appeared in 38 of the 90 answers, the second in 37 and the third in 31. Those three account for 106 of the 431 mentions, about a quarter. Fifteen businesses were named ten times or more.

If you aren't in that top group, you're competing for the long tail.

What the models actually cited

The models cited 802 pages across 113 sites. Two patterns stand out.

Site homepages take about half the citations. Forty-five site homepages account for 391 of the 802 citations. If a business has a well-linked homepage that says what it does and where, the models find it again and again.

City-specific pages punch above their weight. This is the pattern that changed how I think about our site. One business was named in only 2 of the 90 answers. Its Thorold service page was cited 21 times. Another business, named in 10 answers, had a Thorold page cited 21 times too. In both cases the page did the work, not the brand. Other city pages, for Thorold, Niagara Falls and St. Catharines, showed up between 5 and 8 times each.

Two things I expected to matter mattered less than I thought. Pages with pricing in the URL were cited, but only 1 to 3 times each. And directory profiles on sites like HelloDarwin, GoodFirms and DesignRush showed up between 1 and 4 times each. They appear, but the models did not lean on them.

What I changed on our site

Reading this list, the gaps on our side were obvious. A model reading our old homepage would have learned that we were "the digital agency you love" and nothing about where we are. So I changed:

• The homepage title, headline and description, so they say what we do and that we're in Thorold and Niagara.
• Every page title, so each one carries a location.
• Added a page for web design in Thorold, St. Catharines and Niagara Falls, each with its own angle, not the same text with the town swapped.
• Removed numbers on our web development page that I couldn't back up.
• Framed our published packages and pricing for Niagara, in Canadian dollars plus HST.
• Added structured data that says the same thing as the visible text.

None of that guarantees a mention. It just fixes the reasons a model had nothing to find.

What this can't tell you

It's ten questions, one search tool, one snapshot in time, and models change. I didn't test whether being named leads to a phone call, and I didn't measure ad spend, company age or review counts. Treat it as a baseline, not a verdict.

I'll re-run the same questions and publish what changed.

If you want this for your business

I turned the method into a service. The AI visibility audit runs a set of buying questions for your business and town through three AI models and shows you who gets named and which pages the models cite. If you'd like to see what this looks like in a different trade, I ran a smaller version for roofers in three Niagara cities: I asked three AI assistants who the best roofer in Niagara is.