Free tool

AI visibility testFREE

Patients are asking an assistant, not you. Does your name show up in the answer?

This tool does not ask the questions for you, and it does not query any AI service. Based on your specialty and location, it builds fifteen questions of the kind patients actually type; you ask them in the AI assistant you use and mark here what you see. That way you measure the screen a patient really sees, not an estimate from some dashboard. Definitions of the terms are in the glossary.

1. Enter your details

Nothing is sent to a server; everything stays in your browser.

Specialty Focus topic (optional) City Neighborhood or district (optional) Your name or clinic name (optional, for the brand question)

Have the book? Enter the access code from its last page.

2. How to run the test

The method decides the result. Without these four rules, what you measure is your own history.

Do not sign in. Your own account has been talking with you about your practice for months; a test run from that account is flawed from the start. Ask from a browser where you have never signed in. If you have to sign in, use a temporary chat or turn memory off.

Do not use a VPN. Hiding your location does not clean up the test, it distorts it. You are measuring what a patient in your own city sees; the location signal is not noise, it is part of what you measure. Use a VPN only for a separate test: the patient looking from another country.

Write down the conditions. Which assistant, which surface (the app, or the summary inside search results) and which date. Results from different surfaces are not mixed.

Ask an important question three times.Observed For the same question an assistant runs different searches in the background each time, and the answer changes with them. Run each attempt in a new chat. If your name appears in all three attempts, the result is solid; if it appears in only one, treat it as chance and do not mark it.

One measurement is not a result. Ask the same question on another day or in another version and the answer changes. The job of this test is not to give you a score but to show where you stand today. What matters is repeating it once a month and watching your own trend.

The labels on this page show what a statement rests on. Documented an official document or a published study; Observed our own measurements, small sample; Unconfirmed reported, but with no official source.

3. Ask the questions, mark what you see

ChatGPT

Copy each question and ask it in the assistant. Read the answer and look for two things: whether your name appears in the text (a mention) and whether your site is shown as a source (a citation). They are separate; either one can happen without the other.

Full set on: 15 questions.

0questions asked

0answers named you

0answers cited your site

Surface comparison

Mark the same questions separately on each of the three surfaces. The difference between them shows directly where you are missing. Asking the full set on three surfaces means 45 queries; for monthly tracking you can switch on the core set and make the same comparison with 15 queries.

The Tracking ledger is part of the full version. Marking the same questions separately on three surfaces, freezing the question set so you ask the same questions month after month, keeping the whole history and printing a charted report are unlocked by the access code in the handbook. The test and the fifteen questions are open to everyone. Have the book? Enter the access code from its last page above. Get the book.

Open the assistant

After copying a question, open an assistant here in a new tab and paste it. Each one opens in a separate tab, so the marks on this page are not lost.

ChatGPTchatgpt.com Geminigemini.google.com Google SearchAI summary

If you are signed in on the page that opens, switch to a temporary chat or sign out. Your own account's memory distorts the test. The Google search box is here so you can see the AI summary shown above the results; it is a separate surface from the other two and can answer differently. Other assistants are not listed. If your patients use them, you can ask the same questions there; the method does not change.

ChatGPT is a trademark of OpenAI; Gemini and Google are trademarks of Google LLC. Their names are used here only to show which service is meant. This work has no partnership, affiliation or endorsement relationship with the companies named.

How to read the result

If your name never appears, the problem may not be your content. The model may not recognize you as an entity connected with that topic. Publishing more articles does not solve this on its own; your name needs to appear off your site, in independent sources.

If your site is cited but your name does not appear, your content is trusted but your identity is not established. This is a common pattern on physician sites.

The assistant does not search for your question as written.Documented It splits your sentence into several short searches: the city, the name of the procedure, words such as "doctor" or "urologist". Some assistants, ChatGPT in particular, may turn to physician directory and review sites for these searches.Observed If your name does not appear, check two places: does your site have a single, clear page for that city and that procedure, and does your profile on the directory sites list that procedure. A search having been run does not mean being cited either; the assistant may not use what it found.

If both appear, that is still a single measurement; repeat it monthly before drawing conclusions. The terms are explained in the glossary under entity mention and citation, and the full measurement method is in the handbook. To check the article itself, use the Medical Visibility Checker.

See visits from AI assistants separately (GA4)

This test measures whether you appear in the answer; Google Analytics shows how many people actually reach your site from an AI answer. In the default report these visits disappear into "Referral". A custom channel group, set up once, puts them on a line of their own.

Open Admin > Data display > Channel groups and choose Create new channel group. A copy of the default group opens. You need at least Editor access on the property.
Choose Add new channel and name it "AI assistants". As the condition, pick the Source dimension with matches regex and paste this expression:

Move the channel above "Referral" in the order. A visit is counted in the first channel whose definition it matches; left below, these visits still count as referrals.
Save the group. In Acquisition > Traffic acquisition, choose this group as the primary dimension and "AI assistants" appears as its own line. According to Google, a custom group also applies to past data.Documented

A standard property allows two custom groups besides the default one. Update the expression as assistant addresses change. Treat this channel as a floor: some links opened from apps arrive without referrer information and stay in "Direct". The full setup is on Google's custom channel groups help page.