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.
- Mix problem questions, comparison questions and direct recommendation questions.
- Include a couple you expect to lose, so you can see when that changes.
- Include a couple that are genuinely narrow, because that is where movement shows up first.
- Fix the wording and never change it. A changed question resets your trend line.
Eight to twelve is the right size. Enough to see a pattern, small enough that you keep doing it.
Record three signals every run
Named
Did the answer say your business name? This is the outcome that matters most, because a named business is a recommendation.
Cited
Did the answer link to your site as a source? Citation without naming is a real intermediate step and it usually arrives first.
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.