You would never run your shop on "I think production is fine". You count units, track scrap, watch the schedule. Then the marketing conversation turns to AI, ChatGPT this, Google AI Overviews that, and suddenly everyone is trading feelings. Here is the fix: AI visibility is measurable, with a method any owner can run in an hour a month, no subscriptions and no dashboard. You build a panel of the questions your buyers ask, you run them through the engines, you log who gets named, and you watch the trend. Numbers, not vibes.
Quick answer: To measure AI visibility, build a fixed panel of 15 to 25 real buyer questions, run each through ChatGPT, Perplexity, and Google's AI results on a set day each month, and log which companies get named, in what order, and which sources get cited. Track three numbers over time: how often you are mentioned, how often you are the first recommendation, and which sources keep feeding the engines. The trend tells you whether your content, reviews, and citations are working.
What does it mean to measure AI visibility?
AI visibility is the share of relevant AI answers that include your company's name. Not website traffic, not rankings: the plain question of whether ChatGPT, Perplexity, Gemini, Copilot, or Google's AI results name you when a buyer asks the kind of question you exist to answer. It matters because a growing slice of buying research now ends at the answer itself, and the companies named in that answer split the demand while everyone else quietly disappears from consideration.
Classic SEO tools cannot see this. Your analytics show who reached your site, not who asked about your category and got told three competitor names. That blindness is why most companies have no idea they are losing AI-era demand; the loss leaves no trace in any dashboard they own. Measurement replaces that blindness with a baseline and a trend, and it costs an hour a month once the panel exists. The mechanics of getting named are covered in our guide to getting recommended by ChatGPT; this article is about proving whether it is happening.
How do you build a query panel?
A query panel is a fixed list of 15 to 25 questions, phrased the way your buyers actually ask them, that you will run identically every month. Fixed is the operative word: change the questions and you lose the trend. Build it from four buckets:
- Category questions. "Best [what you sell] in [your region or Canada]", "who makes [your product type]", "recommend a supplier for [your specialty]". The head-to-head recommendation queries.
- Problem questions. The symptom-and-situation searches buyers run before they know which product they need. Pull these from what your inside sales people answer every week.
- Comparison questions. "X versus Y", "is [approach A] better than [approach B] for [use case]". Engines love answering these, and they name names.
- Brand checks. "What do you know about [your company]?" and the same for two or three competitors. This reveals what the machines currently believe, including errors worth fixing.
Phrase everything in buyer language, not your internal vocabulary. If your customers say "skid steer door" do not write "loader cab enclosure". If you sell into the US, phrase a few queries the American way too, imperial units and all, for the reasons covered in winning US buyers through AI search.
How do you run the panel and log results consistently?
Same questions, same engines, same day each month, logged in one spreadsheet. The discipline is the method:
- Pick your engines. ChatGPT, Perplexity, and Google's AI results are the practical core; add Gemini or Copilot if your buyers skew that way. Three engines is enough for a real signal.
- Run each query fresh. New chat per query, no personalization from a long conversation. Answers vary between runs, so for your most important five or six queries, run them twice and log both.
- Log four things per answer. Companies named and in what order; whether you appear; what the engine says about you if you do; and which sources it cites, directly visible in Perplexity and Google, worth noting wherever shown.
- Score it simply. Mentioned or not, first recommendation or not. Resist elaborate weighting schemes; they add noise, not insight.
The first month is your baseline, and it is usually uncomfortable reading. That is the point. You cannot manage what you refuse to look at.
What should you track month over month?
Three trend lines, plus one list. Mention rate: the percentage of panel queries where you are named at all. First-recommendation rate: the percentage where you are the lead answer. Source map: the sites the engines keep citing in your category, because those are the places your name needs to live. The list is competitor frequency, which tells you who is actually winning the answer war in your niche, often not who you expect.
Here is the difference between this and the reporting most businesses get:
| Question | Vanity dashboard | AI visibility scorecard |
|---|---|---|
| What it measures | Impressions, sessions, follower counts | Who gets named when buyers ask, tracked monthly |
| What it tells an owner | Something happened | Whether you are in the consideration set or invisible |
| Competitive picture | None | Which competitors the engines trust, and why |
| What to do next | Unclear | Source map and misses point to the exact gaps |
| Cost to produce | A tool subscription and a shrug | An hour a month with a spreadsheet |
| Connection to revenue | Loose at best | Tracks the answers that replace the old first sales call |
Want your baseline run for you?
We'll build the query panel for your category, run it across ChatGPT, Perplexity, and Google's AI results, and hand you the scorecard: who gets named, who gets cited, and the priority fixes. Free, Canada-wide.
Book Free AuditWhat do the results tell you to fix first?
Each failure pattern maps to a specific fix, which is what makes the panel a management tool rather than a curiosity:
- Never mentioned anywhere: a foundation problem. Check whether AI crawlers can read your site at all, the robots.txt and structure issues in our AI crawlers and robots.txt guide, then look at whether any page on your site directly answers the panel questions.
- Mentioned for brand checks but not category queries: the engines know you exist but have no reason to recommend you. That is a content depth and third-party proof gap: buyer guides, case studies, reviews.
- Described incorrectly: an identity problem. Conflicting citations and stale listings are feeding the machines bad data; the cleanup lives in directories and citations.
- Competitors cited from sources you are absent from: the source map is your to-do list. Get present on those directories, publications, and platforms.
How does this fit a weekly scorecard?
AI visibility is a monthly number that sits beside the weekly operational ones. The weekly layer tracks what the business does with demand: answered-call rate, speed to lead, quotes sent, booked calls. The monthly layer tracks whether the demand pipeline itself is growing, and AI visibility is now a core part of that. One without the other misleads: rising visibility with a leaky phone process wastes the wins, and a tight phone process with shrinking visibility runs out of calls to answer.
This pairing is how we report to clients, real numbers on one page, no vanity metrics, because a measurement habit only survives if it stays cheap and honest. Expect slow movement, by the way. Visibility compounds over months, not days; the panel exists to catch the compounding early, prove it to a skeptical owner, and catch regressions before they cost a season.
Frequently asked questions about measuring AI visibility
What is AI visibility?
AI visibility is how often your company gets named when buyers ask AI engines like ChatGPT, Perplexity, or Google's AI results the questions your business exists to answer. It is measured by running a fixed panel of real buyer questions monthly and logging which companies appear, in what order, and from which cited sources. It is the AI-era equivalent of knowing where you rank, except the answer often replaces the visit entirely.
How many queries should be in a panel?
Fifteen to twenty-five, drawn from four buckets: category recommendation questions, problem questions buyers ask before knowing the product, comparison questions, and brand checks on yourself and two or three competitors. Fewer than fifteen is too noisy to trend; many more than twenty-five stops getting done. The panel must stay fixed month to month or the trend breaks.
Which AI engines should we test?
ChatGPT, Perplexity, and Google's AI results are the practical core: they cover the biggest user bases and Perplexity shows its sources, which doubles as a map of where your name needs to live. Add Gemini or Copilot if your buyers work in those ecosystems. Three engines run consistently beats six run occasionally.
How often should we run the panel?
Monthly, same day, same method. AI visibility moves on a scale of weeks to months, so monthly sampling catches every trend that matters without becoming a chore that gets skipped. The exception is after a major change, a site rebuild, a robots.txt fix, a big content wave, when an off-cycle run four to six weeks later shows whether it landed.
Why do AI answers change between runs of the same question?
The engines are probabilistic and answers vary between sessions, which is why the method uses a fixed panel, fresh chats, and repeated runs for the most important queries. What matters is not any single answer but the pattern: a company named in seven answers out of ten is genuinely visible, and one named once in ten is on the bubble. Trend over months, not run to run.
Are there tools that track AI visibility automatically?
A category of monitoring tools is emerging, and automation has its place at scale, but the manual panel remains the best starting point: it costs an hour a month, it forces you to read the actual answers buyers see, and it surfaces the source map and the wording engines use about you, which automated scores flatten away. Start manual, automate once the habit and the panel are stable.
What does AlphaPixels include in an AI visibility audit?
We build the query panel for your category, run it across ChatGPT, Perplexity, and Google's AI results, and deliver the scorecard: your mention rate, who gets named instead of you, what the engines currently say about your company, which sources they cite, and the fixes in priority order. It is free, it works anywhere in Canada, and it becomes your baseline whether or not you work with us afterward.
Related reading
- Directories and Citations Still Matter: The Canadian B2B Edition
- AI Crawlers and robots.txt: Is Your Canadian Company Blocking Its Own Visibility?
- Answer Engine Optimization for Canadian B2B Companies: The 2026 Guide
The bottom line on measuring AI visibility
The companies winning AI-era demand are not guessing; they know their mention rate the way you know your production numbers. One fixed panel of buyer questions, three engines, one spreadsheet, one hour a month. The baseline will probably sting, the source map will hand you the to-do list, and six months of trend will tell you more about your marketing than a year of traffic reports. Measured beats assumed, in the shop and in the answer engines.
If you want the baseline without building the panel yourself, our free AI visibility audit runs it for your category, or book a free fit call with AlphaPixels and we will walk through what the engines say about you today.