Is AI naming your practice? The ten-question test.
Ask four AI assistants, ChatGPT, Perplexity, Gemini and Google's AI Mode, the same ten questions your patients ask, with your city in them, and write down which practices get named. Fifteen minutes and a scorecard tell you whether you exist in AI answers, which competitor does, and which page earned it. It costs nothing.
How to run the test properly
Fresh sessions, signed out where you can, one question per session, everything written down on the same day. The most common way a practice gets a flattering result is asking an assistant it has already told about itself.
Use a private window, or a device that isn't signed in to your accounts. A signed-in assistant can carry your history into its answer, and yours already knows your practice.
One question per fresh session. Don't follow up. The second answer is shaped by the first.
Ask the way a patient would, in plain words, with your city in it. Not "functional medicine SEO Omaha." Ask "what does a first visit with a functional medicine doctor cost in Omaha?"
Write down the practices named and the pages cited. The citations are the more useful half.
Do all forty askings on one day, and put the date at the top of the sheet. These answers move, so the snapshot needs a single date.
AI Overviews are the AI-written summary at the top of some Google results. AI Mode is the chat-style tab beside them. Use AI Mode for the test, because it answers every question, and note when an AI Overview also appears on the ordinary result. Google says optimizing for its AI features is the same work as ordinary search: crawlable pages, accurate business information, helpful content. [Source: Google].
The ten questions patients ask
The mix matters more than the wording: three recommendation questions, three problem questions, two price questions, one comparison and one urgency question, because each kind is answered from a different kind of page. The bracketed words are examples. Put in your own specialty, your city and the problems you see most.
Who is a good [functional medicine] doctor near [city]?I want a doctor who spends more than ten minutes with me in [city]. Who should I see?Recommend three [direct primary care] practices near [city].My child has [common complaint]. What kind of doctor should I see?Is [direct primary care] worth it if I already have insurance?Does [functional medicine] actually work for [condition]?What does a first visit with a [functional medicine] doctor cost in [city]?How much is a [DPC membership] in [city]?[Your practice] vs [competitor], which is better?I need a same-week appointment for my kid in [city]. Who has openings?Questions 4 through 8 are the ones practices skip, and they are where a small practice can win. A new practice rarely displaces the hospital system on “who is good.” Anyone has a chance at “what does a first visit cost,” because it is answered from whichever page states a number plainly, and many practice websites never do. Question 9 is uncomfortable to type. Type it anyway; patients do.
Question 6 is different. Read the answer for claims you would never make. AI answers about health carry medical claims, and an assistant can attach them to your field or to your practice by name. Note anything clinically wrong it says about you, because that is a correction to make on your own pages: the plain, sourced sentence on your site is the version you want it reading next time.
The scorecard
Score each question out of four, one point for each assistant that names your practice. Forty points are available. Then score the same ten questions for the practice that beats you most often, and compare the two totals. The gap between them is your work list.
| # | Question | ChatGPT | Perplexity | Gemini | Google AI Mode | Score | Page cited |
|---|---|---|---|---|---|---|---|
| 1 | Good [specialty] doctor near [city]? | — | — | — | — | 0/4 | — |
| 2 | Doctor who spends more than ten minutes? | — | — | — | — | 0/4 | — |
| 3 | Recommend three near [city] | — | — | — | — | 0/4 | — |
| 4 | My child has [complaint]; what kind of doctor? | — | — | — | — | 0/4 | — |
| 5 | Worth it if I already have insurance? | — | — | — | — | 0/4 | — |
| 6 | Does it work for [condition]? | — | — | — | — | 0/4 | — |
| 7 | First visit cost in [city]? | — | — | — | — | 0/4 | — |
| 8 | Membership cost in [city]? | — | — | — | — | 0/4 | — |
| 9 | [You] vs [competitor]? | — | — | — | — | 0/4 | — |
| 10 | Same-week appointment for my kid? | — | — | — | — | 0/4 | — |
| Total | (40 available) | ? / 40 | |||||
What the total means, roughly
Read the pages that got cited
When an assistant names another practice, it usually shows the page it read. Open it. You are looking at the standard you have to clear, written by someone in your own market, and it is almost never better marketing. It is a plainer page.
What you will usually find
- A number in the first paragraph: the first-visit cost, the monthly membership.
- A heading phrased as the patient's question, not the practice's tagline.
- A table wherever two things are compared: membership against insurance, a first visit against a follow-up.
- The states the practice sees patients in, named one by one, not "several states."
- No throat-clearing before the answer.
For each cited page, note three things: what its first sentence says, whether it contains a number, and whether its heading is a question. Three columns, ten rows. That is a content plan built from evidence rather than from a keyword tool, and it usually points to two or three pages, not twenty.
First sentence
What does its opening sentence state directly?
Contains a number?
Specific visit cost, monthly rate, or patient limits.
Heading is a question?
Phrased as patient intent rather than marketing copy.
What to do with a score of zero
Write two pages, one answering a price question and one answering a problem question, with the direct answer in the first 40 to 60 words and a real number in it. Make your Google listing and your website say identical things. Then run the test again in 60 days with the same ten questions.
Pick the two questions with the clearest decision behind them. Usually a price question ("what does a first visit cost") and a problem question ("my labs are normal and I still feel unwell").
Write one page per question. The first paragraph answers it in 40 to 60 words, in a sentence that still makes sense lifted out on its own. That is the passage an assistant quotes.
Put the real number on the page. The first-visit cost, the membership price, the states you see patients in, named individually. If the number varies, give the range and say what moves it.
Make the code agree with the page, and the page agree with the listing. The hidden fact sheet in your site's code, your visible page and your Google Business Profile should state the same name, address, phone, hours and prices. A contradiction is a reason for a model to reach for someone else.
Check that your site lets the AI crawlers in. OpenAI's ChatGPT search reads the web with a crawler called OAI-SearchBot, and some security plugins and hosts block it. Look at your robots.txt.
Run the test again in 60 days. Same ten questions, same four assistants, same sheet. The same measure every time is what makes a comparison real.
When I did this on a practice site, the page that ended up cited was not the one anyone enjoyed writing. It was the page that said what a first visit costs and what happens at it, and named each state the practice is licensed in instead of saying "several states." The two philosophy pages, the ones we were proudest of, have never been cited by anything. Write the boring page.
What this test can't tell you
It tells you whether you are named today and which pages earned it. It does not predict a position, and it does not call for a separate package per assistant.
- A score is not a ranking. Google says nobody can guarantee a number one position in its results and to be wary of anyone who claims to. [Source: Google]. The same caution applies to an AI answer; nobody controls it.
- You don't need a separate “AI” service. Google's guidance says its AI features draw on the same things ordinary search does. We have found no evidence the underlying work differs between ChatGPT, Gemini and Perplexity; it is the same records, pages and reviews. What's hype, and what isn't.
- Vendor “AI visibility scores” are direction, not fact. They move for the same reason your sheet does. Read the trend over months, not one number.
- The comparison question runs on reviews, and reviews have rules. Since October 21, 2024, an FTC rule bans fake reviews, undisclosed insider reviews and incentives conditioned on a positive review. [Source: FTC]. We are not lawyers; ask yours. How to ask patients for reviews without breaking the rules.
Common questions
ChatGPT, Perplexity, Gemini and Google's AI Mode; each builds its answer a little differently. If patients in your area use another, add a column. The method does not change.
- •Google Search Central: guidance on AI features
- •Google Search Central: do you need an SEO
- •OpenAI: overview of OpenAI crawlers
- •FTC: Consumer Reviews and Testimonials Rule Q&A
- •The ten questions, the scoring and the bands are Ark Visibility's working method, not a published standard. Every practice, question and page on this page is an example; no real practice's results are described.
This guide is general information, not legal or medical advice. AI answers change from one asking to the next; we review this page and date each review at the top.
If you would rather we ran it
See who AI names when patients ask.
The free audit asks these questions from your city, tells you which pages were cited and why, and puts the fix in writing, within two business days.