Privacy: nothing is uploaded. Your practice details, prompts and results are processed locally in your browser, and closing the tab clears them.

I built this for dental practice owners and office managers. I’m Ram Kr Shukla, and I run SEO for 9 US dental practices. If you’d rather hand the whole job off, that’s part of my dental SEO service.

Patients don’t only search Google for a dentist any more. Some ask ChatGPT or Perplexity and get three or four practice names back. Was yours one of them? Nothing in Google Analytics or Search Console will tell you. So this page gives you the questions to ask, a sheet to record the answers in, and a score you can compare month to month.

It’s the hands-on half of a pair. The reasoning (why patient-style questions, how to follow the sources, which fixes usually come first) lives in the AI visibility audit guide for dental practices. Here you just do the work.

How to Use the Prompt Kit

Seven steps. Steps 1 to 3 take about ten minutes. Step 5 is the long one.

1. Copy your details from your Google Business Profile. Search your practice name on Google while signed in to the account that manages your profile. Note the services you list (Google explains the Edit services screen in its services help page), your hours and any attributes such as wheelchair access. Use the same wording in the form. You’re testing the practice you describe to Google, so the two should match.
2. Fill in the form and generate your prompts. Add the neighbourhoods and nearby towns your patients come from, the insurance plans you take, the languages your team speaks and up to five competitors you already lose patients to. Pick 20, 25 or 30 prompts and click Generate prompts. The same details always produce the same prompts, which is what makes next month’s run comparable.
3. Download or print the tracking sheet. Click Download tracking sheet (CSV) and open it in Excel or Google Sheets, or click Print sheet and fill it in by hand. Each assistant gets six columns: mentioned, position, sources cited, competitors named, wrong info and notes.
4. Set up a clean test. Open a private or incognito window and use each assistant logged out where it lets you. If you have to sign in to ChatGPT, switch memory off for the test (OpenAI’s memory FAQ shows where). Location changes answers too. OpenAI says ChatGPT search uses a general location based on your IP address (ChatGPT search help), and Google says Gemini uses your general area, or your precise location if you allow it (Gemini Apps Privacy Hub). Test from the practice or from where your patients live, and write down where you were.
5. Run each prompt in a fresh chat. Paste one prompt, read the whole answer and fill in that row before you ask the next one. Don’t press regenerate: a patient sees the first answer, so that’s the one you record. For Google AI Mode, use the AI Mode tab on Google Search in the same private window. Position is where your practice sits among the practices named, ignoring directories and review sites. Under sources, write the site names the answer links to.
6. Enter your results or import the sheet. Back on this page, pick an assistant tab and fill in the rows, or upload the filled CSV under Import a filled sheet. The tool finds columns by their names, so a file saved from Excel or Google Sheets works.
7. Score it and fix the first gap. Click Score my results and start with the first gap on the list. Next month, import this month’s file and run the same prompts again. Budget roughly a minute per answer: 25 prompts across 5 assistants is about two hours, so it’s worth splitting between two people.

What the prompts cover

The kit spreads your prompts across nine kinds of question, weighted toward the ones that bring new patients: finding a dentist, specific treatments, emergencies, cost and insurance, nervous patients, family and kids, language and access, comparing practices, and questions that use your name. Your neighbourhoods rotate through them, so you see whether you show up for Harbor Point as well as for the city. Prompts for a feature only appear if you ticked it. There’s no point testing “dentist open Saturdays” if you aren’t.

How to Read Your Score and Gaps

The score only counts prompts that don’t contain your practice name. Why? Ask “is Acme Dental Studio good?” and the answer will mention Acme every time, so those prompts are reported separately as recognition: did the assistant know you, and did it get anything wrong?

Every other answer earns points. Named first is 100, second 85, third 70, fourth or fifth 50, lower than that 30. Named in an answer that isn’t a ranked list is 60. Not named is 0. The visibility score is the average, and the Named % beside it is the plain share of answers that included you. Both are this kit’s own yardstick, built for comparing you with you.

The gaps list is ordered by what to fix first:

• Wrong information about you. An old address or wrong hours sends patients elsewhere, even when you’re named. Fix the source the answer cited.
• Not recognised by name. The assistant couldn’t find your practice when asked directly, which usually points to inconsistent name, address and phone details across your listings.
• One assistant far behind the rest. Each comes with a check from that company’s own documentation, such as whether your robots.txt blocks its crawler.
• Weak question types. Named for implants but never for emergencies? The gap tells you which pages and profile fields to look at.
• A competitor named where you weren’t. Reopen those answers and read what they cite for that practice.
• Your site never cited. If you entered your website and it never appears in the sources, assistants are describing you from other people’s pages.

Worked example: the sample practice

Click Load sample data and the kit fills in an invented practice, Acme Dental Studio in Acmeville, OH, with invented results for all five assistants. Here’s how to read what comes back.

• Score 27, named in 35% of answers (46 of 130). That’s the baseline, and on its own it means little. Next month’s number, set beside it, is what tells you something.
• Gap 1 is one wrong answer. Copilot gave an old address for the hours and address prompt. It’s a single row, but it’s the detail a patient acts on, so it goes first.
• Copilot named Acme in 0 of 26 answers and ChatGPT in 4 of 26. Gemini and Google AI Mode both managed 62%. A split that wide points at assistant-specific basics (Bing Webmaster Tools and Bing Places for Copilot, robots.txt for ChatGPT) before anyone rewrites a page.
• Emergency questions: 0 of 15. Acme ticked same-day emergency visits, yet no assistant picked it up. Not one. What to check is the Business Profile hours and a plain page about emergency appointments.
• Nervous patients was the strongest type, at 70%. Leave what’s working alone.
• Sample Smiles Dental turned up in 63 answers that left Acme out. The sources cited in those answers are the next reading list.

The fixes themselves, from Business Profile to treatment pages to schema, are set out in order in AI SEO for dentists. If a gap here doesn’t make sense, the audit guide explains how to trace an answer back to the pages behind it.

What the Assistants Say About Where Answers Come From

None of these companies publishes a list of what makes them recommend one dentist over another. But their documentation does say a few useful things, and each one turns into something you can check.

So anyone who guarantees your practice a spot in these answers is guessing. The docs don’t support it, and the answers shift between sessions anyway. For a wider look at which kinds of sources get named across industries, see which brands ChatGPT, Perplexity and Gemini cite.

Limits of This Kit

• It doesn’t ask the assistants for you. That’s deliberate. Automated checkers run the questions by software, and what they get back can differ from what a patient sees in the app, signed out, in your town. By hand, you read the real answer.
• One run is a snapshot. The same prompt can return a different list an hour later, so give it three months of results before you react to small moves.
• Location, accounts and settings change answers. Write down where you tested and keep it the same each month.
• The prompts are templates built from your details. Real patients phrase things their own way, which is why there’s a box to add questions patients have actually asked your front desk.
• The score isn’t an industry benchmark. Don’t compare it with numbers from paid tools, which each measure differently.
• It can’t tell you why an assistant picked a competitor. The sources column is your best clue.
• Nothing is saved between visits, so keep the CSV. Import it next month and you get back the same prompts and your old results.