AI Chatbot for Dentists: Fee, Emergency and Booking Rules to Set

Give a dental chatbot a short approved fact sheet and three sets of rules: fixed fee wording that never estimates treatment, a clinician-approved emergency message shown whenever a patient mentions pain, swelling, bleeding or injury, and booking limits that stop it confirming anything it cannot see. Hand over to staff whenever those rules are not enough.

The rules matter more than the software. A chatbot with good rules and a modest tool will serve patients better than a clever tool with none. The failures that do most damage in dental chat come from three places: a made-up price, a reassuring answer to a painful symptom, and a promised appointment that does not exist.

Follow me on Instagram@sagnikteaches

Start from an approved fact sheet, not from the tool

Before you open any chatbot builder, write the facts the bot may use. One page, approved by the practice owner and, for clinical wording, the clinical lead. Everything else is out of bounds. Illustrative:

Connect on LinkedInSagnik Bhattacharya

PRACTICE FACT SHEET (approved [date], review every quarter)
Opening hours: [days and times]. Closed on public holidays unless the website says otherwise.
New patients: accepted for [private / mixed] care, as the practice policy states. First visit: examination, about 30 minutes.
Fixed fees: new-patient examination [fee]; hygiene appointment [fee]; emergency appointment [fee].
Payment: card, bank transfer. Finance is offered by [provider] on treatment plans over [amount] subject to the provider's terms and checks.
Cancellation: 24 hours' notice; a fee may apply for missed appointments.
Emergency line: [number], answered [hours]. Outside those hours: [approved wording].
Do not state: treatment prices, success rates, how long treatment takes, anything about drugs, x-rays or what is safe.

Notice the last line. Telling the bot what it must not say is as useful as listing what it may. Keep the fact sheet in one place that reception also uses, so the bot and the team give the same answer.

Subscribe on YouTube@sagnikteaches

Fee rules: fixed prices only, and a standard disclaimer

Patients ask about money first. A dental chat should answer three kinds of fee question differently.

The second row is where most chatbots go wrong. A chat trained on general knowledge will happily offer a typical price range for a crown, and the patient will remember it as a quote. Even with "approximate" in the sentence, that number becomes the expectation on the day. The rule is simple: if the price depends on a clinical finding, the chat does not quote it.

Finance deserves extra care. Offers of credit are regulated in many places, and the exact wording often has to follow the lender's rules. The chat should only repeat wording your finance provider has approved, and never say a patient will be accepted. A safe line: "We offer finance through [provider] on larger treatment plans, subject to their checks. Reception can explain how it works." That is as far as it should go. If you want help drafting the fee sheet and finance wording themselves, the tutorial on admin jobs dentists can give ChatGPT without patient data covers that kind of drafting.

Emergency rules: approved wording, no triage

The chat must not decide what is an emergency. It should recognise that a message might be one, show the practice's approved emergency wording, and offer a human. The words that trigger it should be broad: pain, swelling, bleeding, knocked out, broken, cracked, abscess, can't open, can't swallow, fever, accident, injury, child.

Because the wording is a clinical matter, it should be written or approved by the dentist who leads clinical governance, not by a marketer or the chatbot vendor. Illustrative structure, with placeholders for approved text:

If you are in pain or have had an injury, please phone us now on [number] (open [hours]). Outside those hours, [approved out-of-hours arrangement].
If you have swelling of the face or neck that is spreading, difficulty breathing or swallowing, or bleeding that will not stop, call the emergency services now.
I'm an AI assistant and cannot give advice about symptoms.

Treat the second paragraph as an example only. Whether and how to name those specific warning signs is a clinical decision. Your clinical lead may want different or shorter wording, and some practices prefer to give only the phone number. The point is that someone qualified signs the text off, and the chat repeats it unchanged. Put the date of approval at the top of the fact sheet and revisit it every six months.

Two practical details. First, the message should be the same whether the practice is open or closed, with only the line about hours changing. Second, the chat should not ask for symptom details before showing it. A patient with a swollen face does not want to answer questions first.

Booking rules: what the chat can confirm and what it cannot

There are two kinds of booking chatbot, and which you have decides the rules. A request-taker collects details and passes them to reception. A direct booker writes into your diary. Be honest about which one you have.

• Request-taker. The chat collects name, phone, whether the patient is new, the reason in their own words, and preferred days. It says "reception will confirm" and never states a time. This works with any chatbot and any practice software.
• Direct booker. The chat sees real availability and creates the appointment. This needs a proper integration with your practice management system. Weave's AI Receptionist page, for example, says automatic booking and rescheduling work "for qualified Dental PMS only"; for everyone else it captures appointment requests for the team to review and schedule. Ask any vendor for the list of systems it writes back to, and whether yours is on it, before you plan around instant booking.

Whichever you use, some bookings should never be made by a chatbot: emergencies, anyone who mentions pain or swelling, a child or an accompanied adult with specific needs if your policy says so, and anything involving sedation. For those the chat takes a call-back request and hands over to a person during opening hours. The related tutorial on voice agent contract terms for dentists covers what to negotiate if you choose a phone agent rather than a web chat.

Escalation rules: when it must hand over to staff

Write the hand-over triggers down as a list. A chatbot cannot use judgement; it can follow a rule. Include at least these.

1. Any health concern: pain, swelling, bleeding, injury, a lost filling or crown, a child in distress.
2. Any complaint or expression of anger, including a hint of legal action.
3. Any request about money beyond the fact sheet: negotiation, refunds, insurance cover, finance approval.
4. Any request for records, results, x-rays or a letter about a named patient.
5. Any request the bot does not understand twice in a row.
6. Any request to speak to a person. This should always work, immediately.

The hand-over itself has two modes. During opening hours: offer a live transfer or a call-back within a stated time, such as 30 minutes. Outside hours: show the emergency wording, then take a call-back request for the next working morning. The bot should never leave a patient in a loop of questions. If it has asked "can you rephrase?" once, it hands over on the second failure.

A worked example: the rules in action over a weekend

Illustrative, for a three-dentist practice that opens Monday to Saturday morning. A Saturday-night chat session runs like this.

Message 1. "How much is a check-up for me and my husband?" The chat quotes the approved fee per person, notes that new patients start with an examination, and offers two request slots. No clinical content, within rules.

Message 2. "Do you do Invisalign and how much?" The chat says it cannot quote treatment prices and that the dentist will discuss options and costs at an examination. It does not say yes or no on the specific brand, since that depends on whether the practice offers it and the fact sheet is silent. It offers a consultation request.

Message 3. "My filling fell out and it hurts a bit, can it wait until Monday?" The chat shows the emergency wording, gives the number and the hours, and does not say it can wait. It offers a call-back at opening on Monday. The inbox summary is marked URGENT: pain.

On Monday morning, reception phones the third patient first. That is what the rules were written to produce: a human sees the clinical message in the first five minutes of the day, not at the end of a pile of bookings.

The never-say list

A short list of things the chat must never produce is easier to test than a long list of allowed things. Write yours, and paste it into the instructions.

• Any treatment price, range or "from" figure that is not on the fact sheet.
• "It's probably fine", "that's normal" or "it can wait".
• Any promise that a patient will get an appointment on a date, be accepted for finance, or be treated painlessly.
• Any statement about success rates, longevity of treatment or safety.
• Any comment on another practice, dentist or treatment advice a patient received.
• Any request for a card number, password or full date of birth in the chat.

The last point is easy to forget. A chat window is not a payment page, and a patient who types card details into it has created a data problem for you. Add a line that tells patients not to share payment details in chat and sends them to reception.

Instructions you can paste, then adapt

You are the website assistant for [Practice name], an AI. You are not a dentist and you cannot give clinical advice.
Start every chat with: "Hello, I'm [Practice name]'s AI assistant. I can help with opening hours, fees for standard appointments and booking requests."

1. Use only the fact sheet below. If something is not there, say you don't know and offer a call-back.
2. Quote only the fixed fees on the sheet. For any treatment, say the dentist will give a written plan and cost after an examination.
3. If the message mentions pain, swelling, bleeding, injury, a lost filling or crown, or a child who is unwell, show the EMERGENCY MESSAGE and offer a call-back. Do not ask for symptoms.
4. Take booking requests only: name, phone, new or existing patient, reason in their own words, preferred days. Say reception will confirm. Never state a time.
5. Hand over to staff for complaints, money beyond the fact sheet, records and results, and anyone who asks for a person.
6. Never ask for card numbers or full dates of birth.
7. Keep replies under 60 words.

Testing the rules before patients see them

Write 30 messages. Ten should be ordinary and easy, ten should mention symptoms in awkward ways, five should push on price, and five should try to talk the bot out of its rules. Score each against a fixed expected result, as in the table below, and fix the instructions until all 30 pass.

Run the set again whenever you change the instructions, add a service or when the vendor updates its model. Keep the results with dates. A single failure on a symptom message means the chat does not go live yet.

Costs, tools and what to ask a vendor

A small practice can run a rules-based website chat for little. Reference points at list price as of late 2026: Tidio's Lyro gives a one-off 50 free AI conversations and paid Lyro plans start at $32.50 a month on yearly billing; Chatbase has no free plan, with Hobby at $32 a month billed yearly and its Helpdesk inbox from the Basic plan at $150 a month ($120 yearly). Check the current pages before you decide, because pricing changes. Your practice management supplier may also offer a built-in chat, so ask them first.

Questions to ask any vendor: Which practice systems do you write back to? Can I edit the instructions myself? Can staff take over a live chat? Where is chat data stored and for how long? Is patient content used to train your models? Will you sign the data agreement my rules require? If you have patients in the EU, the AI Act's transparency rules have applied since 2 August 2026: people must be told they are talking to an AI unless it is obvious. The duty sits mainly with the chatbot's provider, but the opening line above is the simplest way to be sure it is met. The tutorial on what to check before hiring an AI agency has more on vendor due diligence, and AI receptionist disclosure and data protection for dentists covers the privacy side.

Insurance, children and other questions the fact sheet must cover

Three more question types come up so often that they deserve a rule each. Decide them before launch.

Insurance and plans. "Do you take my insurance?" invites a guess. The safe answer names what the practice does and does not accept, in one sentence from the fact sheet, and adds "reception can check your cover". The chat should never say a treatment will be covered, because that depends on the policy and the plan. If your practice has its own membership plan, add its name and monthly fee only if the terms are fixed and approved.

Children and guardians. Parents write on behalf of a child, often about pain. Treat any mention of a child with pain, injury or a knocked tooth as an emergency trigger. For routine booking, collect the child's first name and age and the parent's phone number, and let reception confirm who may consent. The chat should not discuss consent rules.

Accessibility and language. A patient may write in another language, with a screen reader, or in a hurry. Decide whether the chat answers in the language it is addressed in, and test that the emergency wording is accurate when translated. If you cannot check a translation, have the chat reply in your approved language and offer a call-back with an interpreter if you provide one.

An escalation table with timings, filled in

Illustrative for a three-dentist practice. Fill in your own numbers and owners, then keep the table where reception can see it.

The targets are placeholders, not benchmarks. What matters is that they exist, they are achievable on your worst day, and the chat states them truthfully. If the chat says "within 15 minutes" and nobody calls for two hours, the patient trusts the practice less than if there were no chat at all.

A weekly review that takes ten minutes

For the first month, read every transcript that triggered a hand-over, plus a random sample of five that did not. Look for three things: a reply that sounded like advice, a reply that included a number not on the fact sheet, and a patient who asked for a person more than once. Add each failure to the test set and fix the instruction line that allowed it. After four weeks the review can drop to a monthly sample. Keep the transcripts only as long as your retention policy allows, and make sure whoever reads them is someone who would normally see that patient data.

How dental chat fails, and how to notice

The helpful price. A patient screenshots "crowns are typically $800 to $1,200" and brings it to the appointment. Watch for any number in a reply that is not on the fact sheet.

The calm answer to a swollen face. The chat replies with general reassurance because the symptom list was too narrow. Test with unusual phrasing and add the words you find.

The invisible booking. The chat says "booked", but the request never reached the diary because the integration failed silently. Send a test booking each week and check it arrives.

The stale fee. Fees change in January and the fact sheet does not. Put a fee review in the same diary entry as the price list update.

The dead end. A patient asks for a person and the chat keeps offering FAQs. Test that "speak to someone" always works, at any point, in any phrasing.

If you want help drafting the fact sheet and the escalation list for your own practice, that is what my AI implementation consultation covers.

Questions dental practices ask about chatbot rules

Should the chatbot quote treatment prices?

Only fixed, approved prices such as a new-patient examination, and always with a line saying the dentist confirms the plan and cost after an examination. Never let it estimate crowns, implants, orthodontics or anything that depends on a clinical finding.

Can the chatbot say whether something is a dental emergency?

It should not judge. It can show the practice's approved emergency wording to anyone who mentions pain, swelling, bleeding, trauma or a knocked-out tooth, give the emergency number, and offer a call-back. The clinical lead should write and sign off that wording.

Will the chatbot book appointments directly?

Only if it is connected to your practice management system and that integration supports write-back. Otherwise it should collect a request and have reception confirm. Check with the chatbot vendor which systems it supports before promising patients instant booking.

How do we stop it making promises we cannot keep?

Give it a short fact sheet, forbid it from improvising, and test it with leading questions such as 'will you match another practice's price?'. Every reply that sounds like a commitment should be on your never-say list.

Further reads

• AI Scribe for Dentists: Will It Write Back to Your Software? — Whether a scribe writes back to your practice software.
• AI Tools for Dentists With No IT Team: Six to Try in the First Month — Six tools to try in the first month without an IT team.
• AI Notes for Dentists: Hygiene, Restorative and Emergency Visit Templates — Note templates by visit type, once the front desk is sorted.
• Heidi Health for Dentists: Setup, Templates and Patient Consent — Setup, templates and consent for a scribe in the surgery.
• AI SEO for Dentists: Getting Found When Patients Ask ChatGPT — Get found when patients ask ChatGPT, before they reach your chat.
• AI Marketing for Dentists: A 30-Day Content Plan With Claim Checks — Content that sends the right patients to your website.
• Dental Chatbot for FAQs: Answers From an Approved Fact Sheet — How to lock a dental website chatbot to an approved fact sheet so it answers fees, hours and finance without diagnosing patients.
• Dental Patient Financing: Compare Plans and Explain Them With AI — Compare third-party dental finance with in-house plans on terms, fees and payout, then use AI to write clear monthly-cost summaries and scripts.
• Open Dental and AI: Which Add-On Tools Connect and Which Do Not — Open Dental ships almost no AI of its own. Here is how to test whether a phone, notes or insurance add-on really reads and writes your schedule.
• Orthodontic AI Receptionist: Consults and Broken-Bracket Calls — A consult-booking script, a broken-bracket handoff and multi-site rules for an orthodontic practice using an AI receptionist.
• Dental Fee Schedule: Set Fees and Explain Them to Patients With AI — Work out a dental fee schedule from chair-minute cost, price a membership plan, and draft patient wording that explains each fee.
• Dental No-Show Policy: Write It, Word It and Send It With AI — One worked no-show policy, the fee versus deposit choice, and message templates for booking, reminders, a missed visit and a waived fee.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Weave AI Receptionist page (automatic booking for qualified dental practice management systems only), Tidio and Chatbase pricing pages, EU AI Act Service Desk (Article 50 and implementation timeline); checked 4 November 2026.