Paste your clinical plan with the patient's identifying details removed and ask the AI for a plain-English explanation covering what's wrong, the options (including doing nothing), costs, number of visits and risks, and have the dentist check every clinical statement before the patient sees it. The AI drafts the words; the dentist stays responsible for the plan and the consent.
The payoff is a document patients can reread at home. Chairside, a nervous patient often takes in the price and little else, then calls back a week later asking what a "post and core" is. A clear written explanation answers that before they ring, and it helps when they talk the decision over with a partner. The risk is equally clear: AI writes confidently, and a confident sentence about the wrong tooth or an overstated success rate is worse than no letter at all. So the workflow below puts the checks before the convenience.
The six questions every patient explanation has to answer
Before writing any prompt, decide what a good explanation contains. Patients want the answers to six questions, roughly in this order:
- What's wrong? In their words, not yours: "the back tooth on your lower left has a crack running under an old filling".
- What happens if I do nothing? Honest and specific, without scare tactics.
- What are my options? Including the cheaper or less invasive ones you'd advise against, with why.
- What will it cost, and when do I pay?
- How many visits, and how long does it take?
- What will it feel like, and what are the risks?
Every draft should be checked against this list. If one question is missing, the patient will ask it by phone.
Turn the clinical note into safe input
A clinical note contains more than the AI needs. Strip out anything that identifies the patient and anything that doesn't affect the explanation. Here's an illustrative before and after.
Before (as written in the notes): "Mrs [surname], DOB [date], [address]. LL6 (36) cracked distal cusp under large MOD amalgam, sensitive to biting, no spontaneous pain, vital. PA: no periapical changes. Hx bruxism, wears night guard irregularly. Options discussed: onlay vs crown. Pt anxious re needles, prefers fewest visits."
After (what goes into the AI): "Adult patient. Lower left first molar (FDI 36): cracked back corner of the tooth under a large old silver filling. Sore when biting, no constant pain, nerve healthy, X-ray shows no infection at the root. Grinds teeth at night, doesn't always wear their night guard. Options: onlay or crown. Patient is nervous about injections and wants as few visits as possible."
The name, date of birth and address are gone, and so is anything else that could single them out when combined, such as a rare occupation. The clinical facts that shape the explanation are still there. If you use a business plan with no training on your content, and your practice has approved it for patient information, you can be less strict, but the habit of removing names costs nothing. How to anonymise client data before you paste it into AI covers the method in more depth.
A prompt that produces a usable first draft
You are helping a dentist write a treatment explanation for
a patient. Use only the clinical facts below. Do not add
diagnoses, success rates, lifespans or guarantees.
Write for an adult with no dental knowledge, reading age
about 11-12. Use short sentences. Explain any dental word
the first time you use it.
Structure:
1. What we found (2-3 sentences)
2. What could happen if it's left
3. Your options - one short paragraph each, including
pros, cons, number of visits. Put [FEE] where the price
goes; I'll add prices myself.
4. What to expect on the day
5. Questions you might want to ask us
The patient is nervous about injections: mention how we
keep treatment comfortable, without promising it's painless.
Clinical facts:
[paste de-identified note]An illustrative extract from the reply:
What we found
The back tooth on your lower left has a crack in one corner,
underneath an old silver filling. That's why it hurts when
you bite. The nerve inside the tooth is healthy, and the
X-ray shows no infection.
What could happen if it's left
Cracks tend to grow. The corner could break off, or the crack
could reach the nerve, which might then need root canal
treatment or, in the worst case, removal of the tooth.
Option 1: an onlay
An onlay is a custom-made cover for the damaged part of the
tooth. It keeps more of your natural tooth than a crown...This is a solid start. What the dentist would fix: "Cracks tend to grow" is fair, but the dentist may want to say how quickly in this case. The draft also mentioned "most onlays last 10 to 15 years" further down, despite the instruction, so that sentence comes out unless the dentist is happy to stand behind a figure. Invented lifespans and success rates are the most common thing to catch; why AI invents confident details explains the habit.
Letting patients compare options side by side
For bigger decisions, a short comparison table does more than paragraphs. Ask the AI to turn the options section into a table, then check each cell. An illustrative draft for a missing lower back tooth:
| Implant | Bridge | Partial denture | Leave the gap | |
|---|---|---|---|---|
| What it is | A titanium post placed in the jaw, with a crown on top | A false tooth fixed to crowns on the teeth either side | A removable plate carrying a false tooth | No treatment |
| Visits | Several, over a few months while it heals | Usually 2-3 | Usually 3-4 | None |
| Affects other teeth? | No | Yes, the neighbours are shaped for crowns | Clasps rest on neighbouring teeth | Neighbouring teeth may drift or tilt over time |
| Removable? | No | No | Yes, taken out to clean | n/a |
| Cost | [FEE] | [FEE] | [FEE] | $0 now |
Two things to check in any table like this. First, the "leave the gap" column should be as honest as the others; patients trust a table that includes the option you'd least recommend. Second, fees go in by hand from your fee list, because a table that adds up wrongly damages trust faster than a spelling mistake.
Adjusting the explanation for the patient in front of you
The anxious patient
Add one line to the prompt: "The patient is very anxious. Lead with what we'll do to keep them comfortable and how they can stop treatment at any time." The draft should then open the "what to expect" section with the numbing gel, the option of a signal to pause, and the length of the appointment. Check it doesn't promise a pain-free experience.
An illustrative draft of that section for the cracked molar, and the edit it needed:
What to expect on the day
We'll start with numbing gel so you barely notice the injection, and
we'll agree a signal, like raising your left hand, so you can stop us
at any point. The appointment takes about an hour. Once the tooth is
numb you won't feel a thing, and most people are surprised how easy
it was.The first two sentences are exactly right. "You won't feel a thing" is a promise the dentist can't keep, since patients often feel pressure and vibration, and an anxious patient who was told otherwise loses trust at the first sensation. The fix is one honest line: "You'll feel some pressure and hear the equipment, but it shouldn't hurt; if it does, raise your hand and we'll stop." "Most people are surprised" goes too, because the practice has no figure behind it.
A parent deciding for a child
Write to the parent, explain in terms the child could also understand, and include what the child might feel afterwards and how long the numbness lasts. Ask the AI for a separate two-sentence version the parent can read to the child.
For a seven-year-old having a baby molar filled, a usable version reads: "The dentist is going to clean a little hole in one of your back teeth and fill it in, so it stops getting bigger. Your mouth will feel funny and puffy for a while afterwards, and then it goes back to normal." The first AI draft for this said "a tiny needle will make your tooth go to sleep". Many dental teams avoid words like needle, drill and hurt with young children and let the dentist explain the numbing chairside in their own words, so that line came out.
A patient who reads another language better
AI can draft a translation, but a fluent speaker should check it, especially the risks and costs. Once checked, save it as a template for that treatment type. Checking AI translations before customers see them has a quick review method.
A phased plan to spread the cost
Ask the AI to group the treatment into phases by urgency ("Phase 1: needs doing in the next month; Phase 2: within six months") with [FEE] per phase. The dentist decides the phases; the AI only lays them out clearly.
An illustrative layout for a patient with a painful upper tooth, two small cavities and a missing lower tooth, after the dentist's edits:
Phase 1 - the next 2-4 weeks [FEE]
Root canal treatment on the upper right tooth that's been aching
(2 visits), then a temporary filling.
Phase 2 - within 3 months [FEE]
Crown on the same tooth to protect it (2 visits).
Two small fillings on the lower left (1 visit).
Phase 3 - when you're ready [FEE]
Replacing the missing lower tooth. We'll talk through the options
in the separate table.
You pay for each phase when it starts.The AI's first version put the two fillings in Phase 1, because it sorted by how quick each job was rather than how urgent. The dentist moved them: small cavities can wait weeks, but an aching tooth that needs root canal treatment shouldn't. That's the kind of clinical judgement the draft can't make, and the reason the dentist sets the phases before the AI touches them.
Where X-ray AI fits in the chair
A different kind of tool helps with the "what's wrong" part. Radiograph analysis software such as Pearl's Second Opinion and Overjet highlights possible findings on the X-ray, such as decay, calculus and bone loss, with coloured overlays, and both companies promote those overlays as a way to show patients what the dentist sees. For many patients, a highlighted shadow on their own X-ray makes the problem real in a way a description doesn't.
Keep two things in mind. The dentist still makes the diagnosis; the software is a detection aid, and a highlighted area isn't automatically something to treat. And check that any such tool holds the regulatory clearance required where you practise before you rely on it. A sensible pairing is the X-ray overlay chairside, then the AI-drafted written explanation to take home.
If you connect the two, watch what the letter does with the overlay. Say the software highlighted three possible areas of decay and the dentist, looking at the X-ray and the teeth, agreed with one. A draft built from a note that lists "3 areas flagged by software" can come back as "your X-ray showed decay in three teeth", and the patient then wonders why only one is being treated. Put only the dentist's findings into the de-identified note, and if the patient saw the other highlights, add one sentence: "The software also marked two areas that I checked and am happy to watch for now."
Checks the dentist makes before anything reaches the patient
- Tooth identification. Does every tooth named match the chart? AI sometimes converts between notation systems: lower left first molar is 36 in FDI notation and 19 in the Universal system, and a draft that swaps them names the wrong tooth.
- No invented numbers. Success rates, lifespans and healing times appear only if the dentist has chosen them.
- No promises. Words like "painless", "permanent", "guaranteed" and "lifetime" come out.
- Every option you discussed is there, including doing nothing.
- Fees match the fee list and the total adds up.
- Reading level. Short sentences, jargon explained. Checking AI content for plain English shows a two-minute test.
- It reads like your practice, not a brochure.
The notation check sounds pedantic until it fails. Picture a practice that records teeth in FDI notation, where a letter drafted from a note mentioning "36" comes back saying "tooth 19 (lower left first molar)", because the model mixed in the Universal numbering it has seen elsewhere. The description is still correct, so the dentist skims past it. The patient, who was shown tooth 36 on the screen, rings to ask why the letter names a different tooth and whether the plan is for someone else. A line in the prompt ("use FDI tooth numbers only, exactly as given, plus a plain description") stops it recurring.
With a saved prompt and a de-identified note, the whole process takes most dentists five to ten minutes per plan, of which the check is the part that matters. In a practice presenting, say, 12 larger plans a week, that's roughly 90 minutes of dentist time, against perhaps four hours if each letter were written from scratch at 20 minutes. The saved hours matter less than the plans that used to leave the practice with no written explanation at all.
Building a library of approved explanations
Most of a practice's treatment plans fall into perhaps a dozen types: fillings, crowns, onlays, root canal treatment, extractions, gum treatment, implants, bridges, dentures, whitening and a few orthodontic referrals. Writing each explanation from scratch every time wastes the check you've already done. Instead, once a dentist has approved a good explanation, strip out the patient-specific lines and save the rest as a master for that treatment type.
Gum treatment is a good one to start with, because patients often don't realise it's a course rather than a single visit. An illustrative approved opening from a master:
Gum disease is an infection of the gums and the bone that
holds your teeth in place. It's usually painless, which is
why it's easy to miss. Treatment happens in stages: a deep
clean below the gum line, usually over [NUMBER] visits,
then a check [TIMEFRAME] later to see how your gums have
responded. Looking after your gums at home between visits
makes a big difference to how well treatment works.When the AI first drafted this, it added "and it's completely reversible", which isn't true once bone has been lost. The approved master fixes that permanently, so it can't creep back in with the next patient.
Store the masters somewhere the whole team can reach, such as a shared folder or a Claude or ChatGPT Project on a business plan, with the dentist who approved each one and the date. A header like this at the top of each master keeps that information with the text:
MASTER: Onlay (single tooth)
Approved by: [dentist initials] Date: [date] Review by: [date]
Fill in each time: [TOOTH - FDI number + plain description], [FEE],
[NUMBER OF VISITS], anxiety line (yes/no)
Never add: lifespans, success rates, "permanent", "painless"
Changed since last version: removed "lasts 10-15 years" (not our figure)The last line matters more than it looks: when a second dentist wonders why the letter doesn't mention how long an onlay lasts, the answer is already there. Review the masters once a year, or sooner if your fees, materials or techniques change. New dentists joining the practice then start from explanations that already reflect how the practice talks to patients.
Knowing whether the explanations are helping
Case acceptance is tempting to track but a poor measure on its own, because it can rise for the wrong reasons if a letter oversells. Better signals, tracked for a couple of months before and after:
- How many patients phone back asking what their treatment involves.
- Whether questions at the consent appointment get more specific ("will the onlay match my other teeth?" rather than "what is it?").
- How often patients change their mind after starting treatment.
- Any complaint that mentions not understanding what they agreed to.
Reception can keep the first measure with a tally mark on a sheet. An illustrative eight weeks either side of the change: before, 60 larger plans presented and 22 calls back asking what the treatment involved; after, 58 plans and 9 calls. The calls that remained were mostly about payment dates, which tells the practice exactly which paragraph of the master letters to rewrite next.
If those move the right way, keep going. For the wider set of consent and data risks when dental practices use AI, see dental practice AI mistakes.
Questions about AI-drafted treatment explanations
Should patients get the written explanation before or after the consent discussion?
After the chairside conversation and before they're asked to agree. That gives them time to read it at home, talk it over and come back with questions. The written explanation supports consent but doesn't replace the discussion with the dentist, which should still cover risks, alternatives and the patient's own priorities.
Can AI translate a treatment plan for a patient who reads another language better?
It can produce a useful draft, but have a fluent speaker check it before it's sent, especially the risks and costs sections, where a slightly wrong word changes the meaning. Keep the checked translation as a reusable template for that treatment type so you only pay for the check once.
Is it safe to include fees in what I give the AI?
Fees on their own aren't personal data, so including your fee list in the prompt is fine. The risk is the AI rounding, adding up wrongly or inventing a fee for an item you didn't price. Add fees yourself from your fee list, or check every figure in the draft against it before the letter goes out.
Further reads
- Winning Back Lapsed Dental Patients With AI Messages — Messages that bring back patients who left treatment unfinished.
- How Dental Practices Use AI for Recalls and Appointment Reminders — Setting up AI around recalls and reminders once plans are accepted.
- Explaining Lens Options and Eye Test Results With AI Help — How opticians tackle the same explain-the-options problem.
- Writing Pet Owner Emails and Treatment Explanations With AI — Treatment explanations for pet owners, a close cousin of this workflow.
- How Accountants Use AI to Explain Tax to Clients in Plain English — Another profession's approach to plain-English explanations with AI.
- What to Ask Before Buying Any AI Tool for a Medical Practice — Questions to ask before buying any clinical AI tool.
- AI Receptionist vs Front Desk Hire: A Dental Practice Comparison — What a dental front desk really does, where an AI receptionist beats a hire and where it can't, and a worked decision for a practice replacing a leaver.
- 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.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Pearl Second Opinion product page; Overjet patient education pages (checked September 2026).