Being named by an AI assistant is not the same as being chosen. Most practices lose the patient in the ninety seconds after the answer arrives: they open the site, find no address, no dentist names and no hours, and go back to the search box. Here is where the drop-off happens, and what your site must confirm to close it.
A patient asking an assistant for a dentist in their city does not get a list of links. They get a paragraph: two or three practice names, a line of description each, sometimes a note about hours. Every claim in it was assembled from somewhere, and the line about being gentle with nervous patients did not come from your site.
The Journey From AI Answer To Appointment
Six steps sit between the question and the booked appointment, and each fails differently. The drop-off column is where practices lose people.
Swipe the table sideways to see every column.
| Step | What The Patient Is Doing | Where They Drop Off | What Closes It |
|---|---|---|---|
| The question | Typing a local, urgent or treatment question | They never ask, and search a directory instead | Nothing you control, except being answerable |
| The answer | Reading a paragraph with two or three names | Not in it, or named without a reason | A specific, checkable description to be lifted |
| The trust decision | Deciding whether the description sounds real | The line about you is vague, or names a service you dropped | Facts on the page the claim came from |
| The verification click | Opening your website | A homepage with no address, no names and no hours | Those three facts above the fold, in words |
| The booking attempt | Calling, or filling in a form | Nobody answers, or the form asks for six fields | A phone a person picks up, a form asking three things |
| The confirmation | Arranging a specific appointment | The patient calls to ask whether to come at all | Same-day confirmation with the practical details |
Steps four and five are fixable this month. Steps one and two are visibility work, and the drsmiles.ai view is that a mention starts the journey rather than ends it.
Where The Money Actually Leaks
A practice can be named in answer after answer and receive nothing, because the failure sits downstream.
What Makes an AI Recommendation Credible To a Patient
Six things get checked before a patient calls, and checked fast.
- Address. Is this the practice near me? An assistant naming a practice without an address has given them nothing to compare.
- Provider name. Who would actually see me? Patients search the dentist's name before they search yours.
- Review evidence. Has anybody else had a decent experience?
- Hours. Can I go today, and are you open when I finish work?
- Service specifics. Do you do the thing I need, and soon?
- One page that confirms all of it. Facts scattered across five pages read as no detail.
Details Patients Check That You Forget Publishing
Sedation offered. Saturday hours. Walk-ins. Whether you are in the patient's insurance network. Which languages the front desk speaks. AI descriptions fail here in two ways: the detail was accurate once and no longer is, or the assistant filled a gap with a guess.
The Verification Step Most Practices Fail
It is nine in the evening. The patient has an assistant answer in one tab and your site in another. What they see is a full-bleed photo of a smiling team and a button saying Request Appointment. Scroll and they find a form asking for a date of birth and insurance details before anyone replies. No address in the header. No dentist names. No hours. No confirmation the practice exists beyond a logo.
That page answers none of the six checks, so the patient closes the tab and asks again: find someone closer who is open now. The second answer is worse for you, because the assistant now weighs distance and hours over quality.
The homepage is not a summary of your practice for a patient who already trusts you. It is the evidence pack for a patient who does not trust you yet.
What Your Site Must Confirm For a Recommendation To Convert
Every fact a patient wants to check has a home. Put it where it is read.
Swipe the table sideways to see every column.
| Fact a Patient Checks | Where It Should Live | Why That Place |
|---|---|---|
| Street address | Header of every page, plus the footer | Near me questions are the most local |
| Phone number | Header, as text rather than an image | An assistant cannot read digits from a picture, and neither can a patient on the phone |
| Opening hours, with exceptions | Footer of every page and the contact page | Missing hours get read as closed |
| Dentist names | Home page and each treatment page | Patients search the clinician as often as the practice |
| Treatments offered | A page for each, plus a plain list | A list is quotable; a brochure is not |
| Price range or how to get one | A treatment page and one costs page | Cost questions arrive closest to a decision |
| Walk-in and emergency policy | One page, linked from the header | Urgency questions get answered in minutes, not days |
| Reviews | Footer link to the profiles patients write on | Reviews are where the recommendation is confirmed |
| Insurance and payment options | A page titled for it | A buried PDF answers nobody |
Two matter more than the rest. The phone number has to be text, and the address has to be in the header, because the page a patient reads is rarely the one you designed.
Here is the version we would run on a practice site in an afternoon:
- Street address and a clickable phone number in every header.
- Opening hours in the footer of every page, holiday exceptions included.
- Every dentist named on the home page and the page they cover.
- A plain list of your treatments, each linking to its own page.
- One page answering what it costs, with ranges and what changes them.
- One page saying what an emergency means at your practice, and who to call after hours.
- A review link in the footer, pointing at the profiles patients write on.
- Insurance and payment information on a page titled for it, not in a PDF.
- A contact form asking for a name, a phone number and one line of context.
- Someone whose job it is to answer that form during opening hours.
The Reviews Question
Patients treat reviews as the confirmation step for whatever the assistant said. A short, recent, specific set changes the decision; a long, vague set does not.
Two properties do the work. Recency tells the patient the practice still runs as it did. Specificity tells them what they would actually get: a review naming the treatment, the wait or a staff member by name says more than a star rating, which says only that somebody was pleased.
Your replies count too. A calm reply that addresses the substance of a complaint tells a reader how the practice behaves when something goes wrong, which is the uncertainty a patient has when choosing between two practices.
The rules for collecting reviews without breaking platform rules are in why online reviews matter for dental AI visibility. And if the assistant described you in a way you cannot verify, reviews are where you correct it: not by arguing with the assistant, but by making the true description visible on the page the patient lands on.
Measuring The Commercial Effect
Traffic is the wrong instrument. A practice can double its impressions and see nothing change, because the step that matters happens in the patient's head.
Instrument The Booking Conversation
- Ask one question on the phone: how did you hear about us today? People who arrived through an assistant will usually say so and name it, because it is fresh in their mind.
- Add the same question as one optional line on the booking form.
- Log the answers in one place with the date. A week of this tells you whether the channel is worth building.
Then connect it to the answer that started it. You need to know which question was asked when the call arrived, because that is the question you must publish an answer to. If the calls are about Saturday implants, the gap is Saturday implants, not another page about implants.
Be honest about the limit. You will not attribute a specific call to a specific answer: answers vary with wording, model, location and time. What you can measure is the volume of questions arriving and the direction of travel between weeks.
To see which answers your practice is already in, start with the snapshot: ten real patient questions across three engines, about an hour, no card. A finished report is at the sample report, the commercial details are in our pricing and terms FAQ, the wider method is in dental AI visibility, and the trust signals on your own pages are the subject of E-E-A-T for dental content.