Measurement

How do I track whether AI is recommending my business?

There is no dashboard that tells you your position in a conversation. So you build the measurement yourself, and it is simpler than it sounds.

The short answer

Pick eight to twelve real buyer questions, ask them on a fixed schedule, and record three things every time: were you named, were you cited with a link, and who else appeared. Log every reading. The trend is the answer, not any single result.

Why the obvious approaches fail

Asking an assistant about yourself once tells you almost nothing. Answers vary between sessions, between accounts and between days, so a single reading is noise.

Watching only your own brand name is worse than useless, because someone asking for you by name has already decided. The visibility that is worth money is being named when the buyer describes a problem and does not know you exist.

And waiting for a vendor dashboard to solve it means waiting. A consistent method you run yourself beats a precise one you do not have.

Build the question set

Write down the questions a real buyer would type before they knew your business existed. Not keywords, full questions in natural language.

Eight to twelve is the right size. Enough to see a pattern, small enough that you keep doing it.

Record three signals every run

1

Named

Did the answer say your business name? This is the outcome that matters most, because a named business is a recommendation.

2

Cited

Did the answer link to your site as a source? Citation without naming is a real intermediate step and it usually arrives first.

3

Who else

Record every competitor named. This is your live benchmark and their pages will show you the shape of answer that is currently winning.

The two supporting signals

Your analytics will show referral traffic from assistant domains. Treat it as directional rather than complete, because assistants often surface a business without generating a click, and some referrals arrive with no referrer at all. Rising assistant referrals are good news. Flat ones are not proof of failure.

Your server logs are the earliest signal of all. Filter for the AI crawler user agents and watch the volume and the status codes. Crawler activity returning 200 with real content tells you the door is open, which is the precondition for everything else. If that number is zero, nothing downstream can work and you have found your problem.

Keep a trend log

One line per run, in a single file, in a fixed format. Date, question set version, named count, cited count, competitors seen, crawler hits, assistant referrals.

This is the part people skip and it is the part that makes the whole exercise worth doing. A single reading is an anecdote. Thirty readings are a direction, and a direction is what tells you whether to keep going or change approach.

Weekly is the right rhythm for most businesses. Daily is only worth it if something is actively being changed and you want the fast feedback.

Published 07 September 2026 by Carpe DM Strategies. We run this work on our own business every day, which is where the examples come from.

Is there a tool that tracks AI mentions?
Several tools exist and they can save time, but the method matters more than the tool. A fixed question set, asked on a schedule, with results logged consistently, will tell you what you need to know.
How often should I check AI visibility?
Weekly for most businesses. Daily only while you are actively changing something and want fast feedback. The consistency of the schedule matters more than the frequency.
Can I see AI traffic in Google Analytics?
Partly. You can see referrals from assistant domains, but assistants often name a business without producing a click, and some traffic arrives without a referrer, so treat it as directional rather than complete.
What should I do about competitors who keep getting named?
Read the exact pages the assistant cites for them. Those pages show you the shape of answer currently winning that question, which is far more useful than guessing.
Why do I get different answers on different days?
Assistants vary their responses between sessions and accounts, and live retrieval changes as the web changes. That variance is exactly why a trend across many readings is the signal and a single reading is not.

Want a tracker built and running for you?

Book a free 30-minute call. We run this measurement on our own business every day and can set the same thing up for yours.

Book a Free Call