A client forwarded me a proposal last month. Another agency had pitched her, and halfway down the second page was a line saying her AI visibility score was 12.
Twelve out of what?
She asked them. They didn't say.
Maybe you've had a version of this. Or maybe you typed your own category into ChatGPT one evening, watched it recommend two competitors, and felt slightly sick about it.
Either way you're now sitting there thinking, is AI visibility a real thing I can measure, or is someone making numbers up at me?
It's real. You can measure it across ChatGPT, Perplexity, Gemini, and Google AI Overviews. There's just no official score, and I'd be suspicious of anyone who says otherwise. Let me show you what the term means, the five things worth measuring, and a method you can run yourself this week.
What Is AI Visibility? (The Short Answer)
AI visibility is how often, how accurately, and how favourably AI systems such as ChatGPT, Perplexity, Gemini, and Google AI Overviews mention your brand when they answer questions in your category. It's the AI search equivalent of rankings, measured in mentions and citations rather than positions.
That's the version for your board deck. Here's the version I'd give you over coffee.
Say you run a physio clinic in Kitsilano. Somebody opens ChatGPT and types "best physio in Kitsilano". Three clinics come back.
Are you one of them?
That's it. That's AI visibility.
It isn't traffic, which counts the people who reached your website. And it isn't rankings. You can sit at number one on Google and still never get named when someone asks an assistant the same question. I've watched that happen to clients who were, understandably, furious about it.
One thing before we go further, because it'll save you money later. There is no industry-standard AI visibility score. Every tool that hands you a number invented its own recipe using its own list of questions. That's how two vendors give you two different scores in the same week and both of them are being honest.
AI Visibility vs Rankings vs Traffic: What Is Actually Different
People mix these three up constantly, and I don't blame them. But rankings tell you where your page sits in a list. Traffic tells you who showed up. AI visibility tells you whether you were in the answer somebody read instead of visiting your site.
| Search rankings | Website traffic | AI visibility | |
|---|---|---|---|
| What it measures | Where your page sits in a list of links | Visits that reached your site | How often you're named in AI answers, and how well |
| Where you see it | Google Search Console, rank trackers | GA4, server logs | Asking the engines yourself and logging what comes back |
| What good looks like | Top three for terms with buying intent | Steady growth in the right kind of visits | Named in most questions, described correctly, linked as the source |
| Main blind spot | Says nothing about whether an engine names you | Misses everyone who got their answer without clicking | No standard score, and results move between runs |
Here's the part that catches everybody out. Your AI visibility can climb while your website visits fall. I've had to explain that in a monthly report more than once, and it never lands well.
It isn't a contradiction, though. It's the entire reason this metric exists.
Pew Research Center followed 900 US adults' browsing in July 2025. When an AI summary sat at the top of the results, people clicked through to a website 8% of the time. Without one, 15%. And the links inside the summary itself? Those got clicked in 1% of visits.
Being in the answer is the outcome now. The click is a bonus. If that shift is new to you, our guide to GEO vs SEO unpacks it.
The Five Things Worth Measuring
We track a group of BC businesses across about 600 customer questions a month . Five things keep mattering. The rest is decoration.
1. Presence: do you show up at all?
Ask 30 questions. Count how many name you.
Our Kitsilano clinic gets named in 9 of them. Presence is 30%. That's the whole calculation.
Start here, because the other four don't mean much if you were never in the answer. Check each engine separately. Good looks like that number climbing month to month, on a list of questions you haven't quietly rewritten to flatter yourself.
2. Share of voice: how do you compare?
Presence alone tells you nothing about the market. Nine out of 30 could be brilliant or embarrassing. Depends who else is in the room.
So: the clinic gets named 9 times. Its main rival gets named 24. Now you know something.
Write down every competitor that comes up, not just the ones you already worry about. Assistants name businesses our clients have never heard of. Good looks like closing the gap on your two or three real rivals.
3. Accuracy: is what it says about you even true?
Engines will state your address, your services, and your prices with complete confidence. They're wrong more often than you'd think.
Last quarter, just over a third of the BC businesses we track had at least one factual error in how an engine described them . An old address. A service dropped years ago. Pricing from a page that no longer exists.
Cheap to fix, expensive to ignore. Copy every error out word for word. Good looks like zero on the questions that matter.
4. Sentiment: are you the answer or an afterthought?
Getting named isn't the same as getting recommended.
"The best option for a growing clinic" and "other clinics in the area include" are doing very different jobs. Mark each mention one of three ways: recommended, listed, or hedged.
And watch the qualifiers. "May suit smaller budgets" sounds harmless. It isn't. Good looks like moving from listed to recommended on the questions closest to a booking.
5. Citations: is your website the source?
The last one asks whether the engine links your pages as proof for what it just said. That's what generative engine optimization sets out to earn.
Being cited and being named are two separate things, and the gap is wider than I expected. A June 2026 study by Semrush and Growth Memo looked at 3,981 cases where a site turned up in an AI answer. In 61.7% of them the page was used as a source but the brand was never named. Ghost citations, they call them.
Track both. Good looks like being the name in the answer and the link underneath it.
| What to measure | How | What good looks like |
|---|---|---|
| Presence | Count how many of your questions name you | Climbing, on a fixed list of questions |
| Share of voice | Your mentions against every rival named | Gap closing on your two or three real rivals |
| Accuracy | Copy out every error word for word | Zero errors on the questions that matter |
| Sentiment | Mark each mention: recommended, listed, hedged | Moving from listed to recommended |
| Citations | Note when the engine links your pages | Named in the answer and linked underneath |
Can you squash all five into one number? Sure, and your board will probably ask for it. Just be honest about what it is: a weighting somebody chose, ours included. We'd rather report the five separately.
How to Measure Your AI Visibility (A Method You Can Run This Week)
You don't need to buy a thing. An afternoon, a spreadsheet, and the willingness to repeat it the same way every month gets you a baseline you can defend.
- Write 20 to 30 questions a real customer would ask. Mix category ("best physio in Kitsilano"), comparison ("clinic A or clinic B"), and problem-led ("my back has hurt for a month, who do I see"). Write them the way a person talks. Nobody types keywords at an assistant.
- Pick your engines. ChatGPT, Perplexity, Gemini, and Google AI Overviews. Use all four. Doing well on one tells you almost nothing about the other three.
- Ask each question three to five times. Fresh session each time, with memory, history, and personalisation off. Otherwise you're measuring what the engine knows about you, not what a stranger sees.
- Log every run in one row. Question, engine, were you named, who else was, any errors copied word for word, recommended or listed or hedged, and whether your site was linked.
- Turn presence and share of voice into percentages. One set of numbers per engine. Leave the errors as plain text. That list is your first fix-it queue.
- Do it again next month with the same questions. Change as little as you can. Swap questions around and you've broken your own trend line.
Now the bit most vendors leave out. Ask the same question twice and you'll often get two different answers. Not slightly different. Different businesses named.
The 2026 State of AI Search, from AirOps and Kevin Indig, found that only 30% of brands stay visible from one answer to the next. Just 20% survive five runs in a row. Our own tracking agrees: two runs in three come back with a different list of names .
So a screenshot proves nothing. Not the flattering one, not the terrifying one. Ask repeatedly or don't bother asking.
“These are different behavioral systems. Treat them that way.”
Kevin Indig, Growth Advisor, in Semrush and Growth Memo's ghost citations study
He means the engines. In that study, the businesses ChatGPT linked to and the ones Gemini named for the same question barely overlapped. Average the four together and you'll smooth away the thing you needed to see.
Want to watch someone else do this first? Our Canadian AI visibility benchmark runs the exact method above across 50 Canadian brands and all four engines. There are tools too, including Ahrefs Brand Radar and the Semrush AI Visibility Toolkit, but I'm not ranking those here. That comparison comes later this year.
What Counts as a Good AI Visibility Score?
I get asked this constantly, and the honest answer disappoints people. Anyone quoting a universal benchmark is guessing. Good depends on two things: how crowded your category is, and where you started.
In a local or niche market, being named in most of your questions is a strong position. A business that has never touched AI search usually starts somewhere between four and nine percent , and quite often at zero on the comparison questions that matter most.
A national category full of big spenders behaves nothing like that. What tells you something is direction over three months. Going from 6% to 18% on a fixed list of questions is real progress. Hitting 40% once, on a list you wrote last Tuesday, is noise.
For real Canadian numbers instead of an invented target, the 2026 Canadian AI Search Visibility Benchmark has the spread across 50 brands and four engines.
What to Do When Your AI Visibility Is Low
If you've never worked on this, a low number is completely normal. Don't panic. Four things move it:
- Answer the question first. Put the direct answer at the top of the section instead of building up to it. Engines lift passages, not whole pages. For the Google side, see how to show up in AI Overviews. For assistants, see how to rank in ChatGPT.
- Add structured data. Schema tells an engine what a page is instead of making it guess. It's the heart of answer engine optimization, and on most sites it's a quick win sitting untouched.
- Say the same thing everywhere. Your name, address, and service list have to match across your site, your listings, and your profiles. Mismatches are the main reason engines get your details wrong.
- Get other people talking about you. Engines lean hard on what other sites say. Listings, reviews, and roundups like Vancouver AI SEO agency comparisons all feed that.
Rather hand it over? Our generative engine optimization service and answer engine optimization service each cover half of it. AI search optimization runs both as one program with a single scoreboard.
Frequently Asked Questions
What is AI visibility in simple terms?
It's whether AI assistants name your business when someone asks a question in your field. If a customer asks ChatGPT for a recommendation and gets your competitor instead, your AI visibility is low. It covers being named, described correctly, and linked as the source.
How do you measure AI visibility?
Write 20 to 30 questions a real customer would ask, then ask each one three to five times in fresh sessions across ChatGPT, Perplexity, Gemini, and Google AI Overviews. Log whether you were named, who else was, whether the details were right, and whether your site was linked. Repeat monthly with the same questions.
What is a good AI visibility score?
There's no industry standard, so it depends on your category and your starting point. In a local or niche market, being named in most of your questions is strong. Improvement over three months tells you far more than any single number.
Is AI visibility the same as SEO?
No, though they're related and good SEO helps. SEO measures where your pages rank in a list of links. AI visibility measures whether your business appears inside the answer itself. You can hold number one on Google and still never get named by an assistant.
Why do AI engines give different answers each time I ask?
Because the engine builds the answer fresh every time instead of reading a fixed ranking, so the businesses it names shift from run to run. The 2026 State of AI Search found only 30% of brands stay visible from one answer to the next. That's why one screenshot proves nothing.
So, back to that proposal with the 12 on it. My client now knows what to ask back: 12 out of what, on which questions, across which engines, and how many times did you ask? Those four questions beat any score somebody hands you.
Here's what I'd take away. AI visibility is how often, how accurately, and how favourably AI systems name your business when they answer questions in your category. Measure it in five parts, measure it repeatedly, and accept that any composite number, ours included, is a judgement call. None of which is a reason to wait. Whoever writes down a baseline this month is the only one who can prove they improved six months from now.
So, one question before you close this tab. If a customer opened ChatGPT right now and asked for a business like yours, would your name be in the answer?
You can find out this afternoon. It's free.
We track AI visibility for BC businesses across ChatGPT, Perplexity, Gemini, and Google AI Overviews, then do the work to move it. You get the number before the pitch. See it at full scale in our 2026 Canadian AI Search Visibility Benchmark, or what's covered in our AI SEO and search visibility service.
