Winning Back Lapsed Dental Patients With AI Messages

Pull a list of patients with no examination for about two years, clean out anyone who has died, moved or asked not to be contacted, then split the rest by why they probably stopped coming. Use AI to draft a short, specific message for each group, send three or four touches over about six weeks, and have a person phone the most important ones.

Segmenting is what makes this work. A single "we miss you" text to 1,200 people gets ignored by most and irritates some. Patients who left halfway through a treatment plan, patients who were nervous, and families whose dentist has retired need different messages, and some of them need a phone call from a clinician rather than a marketing text at all. AI is good at writing those variants quickly; the judgement about who goes in which group stays with the practice.

Follow me on Instagram@sagnikteaches

Who counts as lapsed, and who doesn't

There's no single definition, so pick windows that suit your recall intervals and stick to them. A workable set for a general practice:

Connect on LinkedInSagnik Bhattacharya

• Overdue (up to six months past their recall date): still a recall problem. Keep them in your normal reminder sequence (how dental practices use AI for recalls and reminders covers that).
• Lapsed (roughly 6 to 30 months past due): the group this tutorial is about. They still think of you as their dentist, even if they haven't been.
• Dormant (beyond about 30 months): many will have moved or registered elsewhere. One final message, then archive in line with your records policy.

Run the report from your practice management system with last examination date, recall due date, last treatment plan status, usual clinician, contact details and any "do not contact" or complaint flags.

Subscribe on YouTube@codingliquids

Clean the list before anyone gets a message

This stage protects patients and your reputation. Remove or hold back:

• Anyone recorded as deceased anywhere in their notes, not just in the status field.
• Patients who asked not to be contacted, or who opted out of messages.
• Patients discharged from the practice, or with an open complaint or unpaid bill. These go to the practice manager, not an automated campaign.
• Patients with a note saying they've registered elsewhere.
• Children, whose messages should go to a linked parent's contact.

An illustrative clean-up for a two-dentist practice: 1,150 patients in the lapsed window; 84 removed for deceased, opt-out, complaint or registered-elsewhere notes; 176 with no valid mobile or email, moved to a letter-only group. That leaves 890 reachable by text or email.

A note on rules. Reminding a current patient about care is usually treated differently from marketing, but a reactivation campaign to people who haven't visited in two years can fall on the marketing side in some places, which may need consent and always needs a clear way to opt out. Check with your data-protection adviser before your first campaign, and follow your professional regulator's rules on advertising, which generally bar misleading claims and pressure.

Five groups and what each needs to hear

The first group deserves particular care. A patient who left with a temporary crown or a half-finished root canal isn't a marketing target; they're a clinical risk. Have the treating dentist, or another dentist, decide who to call and what to say.

Drafting the messages with AI

Write templates, not personalised messages, so no patient data goes into the AI. Your messaging platform fills in names and details. A prompt for the nervous-patient group:

Write 3 text messages (each under 300 characters) from a
dental practice to adult patients who haven't visited for
about two years and are flagged as nervous about treatment.

Goals: make it feel safe to come back; offer a short
"chat first" appointment with no treatment; make replying
easy. Rules: no guilt ("it's been ages"), no warnings about
what might be happening to their teeth, no discounts, no
exclamation marks, no emojis. Use [first name],
[practice name], [booking link]. Include "Reply STOP to
opt out."

An illustrative reply, with the one worth keeping:

Hi [first name], it's [practice name]. If coming back to
the dentist feels like a big step, you're not alone. You
can book a 15-minute chat with no treatment and no
pressure, just to talk things through: [booking link].
Or reply CALL and we'll ring you. Reply STOP to opt out.

Good. Two of the other variants needed fixing. One opened with "We've missed your smile", which many patients find twee. Another said "don't let a small problem become a big one", a mild scare line the prompt had ruled out. Always read every variant against the rules you gave; AI drifts back towards marketing clichés.

For the "dentist has left" group, the AI can turn a new dentist's notes into a short, warm introduction: where they trained, what they enjoy treating, one human detail they're happy to share. Have the dentist approve it. From notes such as "qualified 2014, enjoys treating nervous patients and children, keen cyclist, happy to mention it", an illustrative first draft read:

"Dr [surname] joined us in the spring and is now looking after
patients who used to see Dr [previous dentist]. With over ten years'
experience and a special interest in anxious patients, Dr [surname]
is known for a gentle, unhurried approach..."

The dentist cut "special interest", because in dentistry that phrase can suggest a formal qualification she doesn't hold, and replaced it with "particularly enjoys looking after nervous patients". "Known for" also went: she'd been at the practice five months, and nobody could say what she was known for yet. The cycling line stayed, and was the part patients mentioned when they booked.

For the unfinished-treatment group, use AI only to tidy a letter the clinician has drafted, and watch that the tidying doesn't soften the clinical point. A dentist's draft said: "When you were last with us you had a temporary crown on your lower right back tooth. These aren't designed to last, and I'd like to check it and finish the treatment." The tidied version came back as: "You may wish to consider returning to complete your treatment at a time that suits you." Smoother, and it had removed the only reason the letter existed. The dentist kept her original sentence and accepted only the reformatting of the booking details underneath it. A useful instruction for this group is: "Improve layout and grammar only. Do not change any clinical statement or its urgency."

A six-week sequence that knows when to stop

The closing message does two jobs: it respects patients who aren't coming back, and it cleans your list for next year. Anyone who books at any point drops out of the sequence at once; check your platform actually does this before launch, because a "we'd love to see you" text sent the day after someone's appointment looks careless.

Households need the same care. In one realistic slip, a mother whose three children were all lapsed received four texts in a single morning: one for herself and one for each child, all sent to her number because the children's records were linked to it. She rang to ask whether the practice's system had broken. Before each send, sort the list by mobile number and merge any number that appears more than once into a single family message ("we'd love to see you and [number] others in your family"), so each household gets one text per touch.

Making the phone calls without sounding like a call centre

The day-28 calls convert better than any text, but only if they sound like the practice rather than a sales line. Give reception a short prompt card, not a script to read, and let AI help you write it: describe the group, the goal of the call and three things the caller must never do (pressure, mention what might be wrong with the patient's teeth, or argue about a past bill), and ask for an opening line, two follow-up questions and a polite way to end.

Here's the difference it makes, for a family who haven't been in two years. A weak opening: "Hi, I'm calling from the dental practice because our records show you're overdue for your check-ups." It sounds like a debt reminder, and the parent's guard goes up. A better one: "Hi, it's [name] from [practice]. I'm ringing because we've got some after-school slots coming up and I wondered if it'd help to book the whole family in together?" It offers something useful in the first sentence and makes the decision easy.

Keep calls short, no more than two or three minutes, and log the outcome in the practice software straight away: booked, call back, moved, registered elsewhere, not interested. Those outcomes feed next year's list, and they stop a patient who said "not interested" being rung again next month by a colleague who didn't know.

Handling the replies people actually send

Replies are where reactivation succeeds or fails, and they need a person or a very carefully limited AI. Illustrative replies and the right response:

• "I moved away last year." Thank them, mark the record, stop all messages. No attempt to keep them.
• "I'm terrified of the dentist, that's why I stopped." A person replies, ideally offering a call with a dentist or nurse who's good with nervous patients. Never an automated booking link on its own.
• "Can't afford it right now." Reply kindly with your payment options if you have them, and leave the door open without pressure.
• "I go to another dentist now." Thank them, mark as registered elsewhere, stop messages.
• "Stop texting me." Treat it exactly like STOP, immediately, and apologise.
• "My tooth has been hurting for weeks." Pass to the clinical team the same day; this is an urgent care question, not a booking.

The cost reply is where AI-drafted answers most often go wrong, because the model reaches for a sales close. Asked to answer "Can't afford it right now", one illustrative draft said: "We completely understand! Our flexible payment plans mean cost doesn't have to be a barrier. Book today to secure your spot." Three problems: it argues with the patient, it promises that a plan will make treatment affordable when nobody knows what they need, and "secure your spot" adds pressure. The version reception sent:

"Thanks for letting us know, [first name]. There's no rush. If it
helps later on, a check-up costs [price] and we can spread the cost
of any treatment. Just reply whenever suits you and we'll find a
time. We won't send any more reminders this month."

The last sentence does the most work. It tells the patient the pressure is off, and it commits the practice to something checkable: that patient has to be paused in the sequence, not just sent a nicer reply while the day-21 text goes out anyway.

If replies pile up, AI can help sort them: export the reply text without names or numbers, ask it to put each into one of the categories above and flag anything mentioning pain or distress, then check its flags by hand. An illustrative run on four replies:

Reply 1: "we moved to the coast in march"          -> Moved
Reply 2: "cant do weekdays, any saturdays?"          -> Wants to book
Reply 3: "a bit of my filling came out ages ago,
          been meaning to come"                      -> Wants to book
Reply 4: "not now thanks, money is tight"            -> Cost

Three are right. Reply 3 is the one that matters: a lost piece of filling isn't pain, so it wasn't flagged, but it is a clinical issue that shouldn't wait for the next routine slot. Widening the prompt to "flag any mention of pain, a broken or lost filling or crown, swelling, bleeding or a tooth that feels different" caught it on the next run. Even then, read every reply the AI sorted as "wants to book", since that's the category where a clinical detail is easiest to pass over.

What the numbers might look like for a two-dentist practice

Continuing the illustrative practice with 890 reachable lapsed patients, plus 176 contacted by letter. Suppose, over six weeks:

• around 11% of the reachable group book, about 98 patients, most after the first text or the phone call;
• another 60 or so reply MOVED or say they're registered elsewhere, which cleans the list;
• a dozen opt out;
• the letter group books at a lower rate, perhaps 5%, about nine patients.

If each returning patient's first year is worth around $350 in examinations, hygiene visits and routine treatment, about 107 returners is roughly $37,000 of work, from a campaign costing mainly reception time: perhaps 15 hours for the calls and replies. These figures are only an illustration; measure your own bookings from the campaign, and count only patients who actually attend.

Some of those patients would have come back anyway, so a simple hold-out check tells you what the campaign really did. Before sending, set aside a random tenth of the routine, no-flags group (in this example about 50 patients) and don't message them for the six weeks. If 2 of the 50 book on their own (4%) and 11% of the messaged group book, the campaign's real effect is about seven percentage points, not eleven. That's still worth doing, and it's a more honest number to take to a practice meeting. After the six weeks, the held-back patients simply join the next round. Only hold back the routine group: patients with unfinished treatment or flagged concerns should never be left out to make a measurement tidier.

Winning back lapsed customers with AI-personalised emails covers the email side in more depth, and the barber's version, winning back lapsed clients with automated messages, shows how much simpler it is when there's no clinical dimension.

Keeping them once they come back

A reactivated patient who has a rushed first appointment and no clear plan is likely to drift again. Three habits help:

• Flag returners in the diary so the dentist knows it's their first visit in two years and allows time to talk.
• Send a written summary of anything the dentist recommends, in plain English; explaining treatment plans with AI shows how to produce one in minutes.
• Book the next recall before they leave, then let your normal reminders take over.

Before any of this, make sure patient details only go into tools your practice has approved for them; the patient data confidentiality checklist covers what to check.

Further reads

• Dental Practice AI Mistakes: Consent, Data, and Over-Automation — Consent and data mistakes to avoid across all your patient messaging.
• How to Reduce No-Shows With AI Reminders and Automatic Rebooking — Keeping reactivated patients from missing the appointment they just booked.
• AI Receptionist vs Front Desk Hire: A Dental Practice Comparison — Whether AI or a person should make the follow-up calls.
• How to Write Sales Call Scripts With AI That Don't Sound Scripted — Call scripts for the phone stage that don't sound read out.
• AI Service Reminders That Bring Garage Customers Back — How garages bring customers back, with ideas that transfer.
• Email Newsletter QA Checklist Before You Hit Send — A final check before a bulk message goes to hundreds of patients.
• How Independent Opticians Can Use AI for Recalls and Bookings — Four kinds of optician recall, the wording that gets them booked, an assistant for booking questions, and the figures to track, for independent practices.
• Which Dental Front Desk Tasks Can AI Take Over? — Task by task: which dental reception jobs AI handles well, what each looks like in practice, and the calls, forms and conversations that stay with your team.
• How a Clinic Can Use AI to Cut Missed Appointments — One physiotherapy clinic's eight weeks: finding where missed appointments cluster, timing reminders per appointment type, and refilling freed slots.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general guidance only; no vendor-specific figures used. Check your practice software's reporting and your messaging platform's segmentation features.