Measurement & Reporting · 4 min read · September 24, 2026 · Hafsa Tariq, gaash.ai
Does the AI recommend your dental practice? A 30-minute check, and what to do with the result
You can find out in about half an hour whether the assistants name your practice. Write six questions the way patients ask them, put each to ChatGPT, Claude, Gemini and Perplexity in a fresh chat, and write down who is named and what is said about you. It is a first look, not a measurement: answers vary between runs, so a monthly panel is what tells you whether anything changed. But half an hour is enough to know where to start.
The six questions
Fill in the brackets once and keep the wording. Write the town or neighborhood the way patients say it; "near me" means wherever you happen to be sitting.
- "dentist in [town] accepting new patients"
- "emergency dentist in [town] open on [day] morning"
- "dentist in [neighborhood] who takes [a plan you bill]"
- "[a treatment you want more of] in [town], who is good"
- "[your practice] or [a nearby practice] for [a treatment]"
- "is [your practice] in [town] taking new patients, and what are the hours"
How to run them in half an hour
- Open a new chat for each question, signed out or with memory switched off where the assistant allows it.
- Ask each question once in each of the four assistants, with web search on where it is offered. Twenty-four answers take about twenty minutes.
- Paste every answer into a document with the assistant's name.
- For each answer, write down: were you named, were you recommended or only listed, who else was named, and every fact stated about you.
One run per question is the shortcut that makes this take half an hour, and it is also its limit: the same question can name different practices on the next run. Treat what you find as a direction, and see AI visibility for dentists for the monthly panel that turns it into a measurement.
The facts to verify in each answer
| Fact in the answer | Check it against | Where a wrong version can live |
|---|---|---|
| Accepting new patients | What your front desk would say today | An old page, a directory, a plan's provider list |
| Hours and emergency care | This week's schedule | A listing never updated after a change |
| Plans billed | Your current list | A PDF from years ago, a directory profile |
| The dentists | Who works there now | A team page, a booking profile, a dentist's old listing |
| The address and entrance | Your door | A map pin, a building's other address |
| Treatments offered | What you do | A services list that was never pruned |
Where an answer shows its sources, open them; the page behind a wrong fact is where the fix goes. When AI gets facts about your business wrong covers the order.
What each result points to
| What you found | What it points at | Where to start |
|---|---|---|
| Never named, on any assistant | Access, or a practice the answers cannot place | The crawler access check, then one name, address and phone everywhere |
| Named only when asked by name | Pages that describe you but not what patients filter by | A dated new-patients line, the plans list, treatment pages |
| Named, with a wrong fact | A page or listing that still states it | The source the answer cites, then your own pages |
| Named by one assistant, missing on another | What that assistant's crawler can reach | robots.txt and the security settings in front of the site |
| A nearby practice named instead | What its pages state that yours do not | Its page for that question, read against yours |
In the probe that accompanied our audit of 9,483 small business websites, Claude named 14.9% of 9,240 businesses on one question about their own trade and town. Across 40 trades, being left out was the ordinary result, so a thin first check is a starting point rather than bad news.
After the half hour
Fix the wrong facts first, because they turn patients away today. Then work down the table from the top, one row at a time, and set up the monthly panel so you can see whether the answers follow. The dental practices overview lists what the practice's pages should state, and treatment pages covers the pages for the work you want more of.
Common questions
Is one run of each question enough?
For a first look, yes; for a measurement, no. Answers vary between runs, so one run shows a direction. The monthly panel asks each question three times on each assistant and compares month with month.
Which assistants should we check?
ChatGPT, Claude, Gemini and Perplexity, with web search on where it is offered. They fetch through different crawlers and write different answers, so a practice can be named by one and missing on another.
Should staff run the check from the practice's computers?
Use a fresh chat and, where possible, a signed-out session or one with memory off, so the answer is closer to what a new patient sees. Write the town into every question rather than relying on where the computer is.
Can someone run it for us?
The free visibility check is the same idea with one assistant, Claude, run by a person, at $0, with a written result within 48 hours. It tells you whether Claude names you and what on your site is most likely in the way.
Sources
The probe figure is this site's own, from the audit named below, and covers all 40 trades with no dental cut.
Sources: our audit of 9,483 small-business sites, with its method · how the visibility score is measured
Read on