AI Chatbot for Opticians: Appointments, Repairs and Lens Questions

An AI chatbot can answer an optician's routine questions from an approved fact sheet: opening hours, how to book an eye test, what to bring, repair and collection status once staff have updated it, and how to start a contact-lens reorder. Anything about eye symptoms, prescriptions or clinical judgement goes straight to a person.

The simple rule behind this is that the chatbot may report what the practice has already decided, and may never decide anything itself. Opening hours are decided. "Is this flashing in my eye serious?" is not, and no wording of the bot's reply can make it safe. If you hold that line, a chatbot takes a large share of repetitive questions without touching a patient's health.

Follow me on Instagram@sagnikteaches

Sorting every enquiry into three piles

Start by looking at what customers actually ask. Take a fortnight of phone notes, emails and walk-in questions and sort them into three groups. The table shows an illustrative split; your own practice will move some lines between the columns.

Connect on LinkedInSagnik Bhattacharya

The third column is the important one. It must be short, absolute and tested. Everything in it triggers the same fixed behaviour, with no clever wording from the model.

Subscribe on YouTube@sagnikteaches

Write the approved fact sheet first

A chatbot is only as good as the document it answers from. Write a fact sheet of one to three pages before you choose any tool. It holds the practice's own answers, in plain words, each with a date and an owner. A suggested layout:

• Practice basics. Hours, contact route, accessibility, parking, payment methods accepted.
• Eye tests. How to book, how long they take, what to bring (current glasses or lenses, a list of medicines, any previous prescription), whether children and people needing extra support are catered for, and the cancellation rule.
• Repairs and adjustments. What can be done on the spot, what is sent away, usual turnaround, how customers will be told it is ready.
• Contact lenses. How to start a reorder, what staff will check, how long orders take, what aftercare appointments are required. Stick to your own policy and wording.
• Frames and lenses. Brief descriptions of lens options as your own dispensers explain them, without claims you cannot support, and how to book a dispensing appointment.
• What the chatbot must hand over. The list from the third column above, with the exact message to show.

Keep it factual. If you do not state a price, the bot cannot state one, and a bot that says "ask a member of the team" is better than one that guesses. The method matches the one in building the FAQ your chatbot needs before launch, which is worth reading for how to word each answer.

Booking an eye test

This is the highest-value job. A customer who can book at 9 pm without waiting for the shop to open is a customer who books. There are two set-ups, and which you choose depends on your booking software.

• Request only. The bot collects the customer's name, contact details, preferred days and any needs (for example a child, a wheelchair, a hearing difficulty). It sends an enquiry to the team, who confirm. This needs no connection to your booking system and carries little risk.
• Live booking. The bot shows free slots from your booking system and reserves one. This is more convenient and needs a chatbot tool that connects to the diary. Check the connection and test thoroughly, because a double-booked test slot costs you a lot of goodwill.

For either, write the questions the bot asks in a fixed order, and keep them short: type of appointment (eye test, contact-lens check, dispensing), preferred day, name, phone or email, and a free-text box for anything the practice should know. If the customer writes something about symptoms in that box, the bot's hand-over rule applies before the booking is confirmed.

Repair status without making things up

Customers ask "Are my glasses ready?" far more often than the practice expects. The bot can answer only if it has a source of truth. Choose one of two honest approaches:

• Staff-updated status. Staff mark each job as received, sent away, back in practice or ready to collect in a sheet or a booking tool the bot can read. The bot reports the status in plain words and adds the usual turnaround from the fact sheet.
• No live data. The bot takes the customer's name and job reference, passes the enquiry to staff, and says someone will reply within a stated time. This feels slower but is always accurate.

What the bot must never do is estimate. "It should be ready by Friday" from a model that does not know is worse than "I have passed your question to the team." The same status logic applies to online orders, which is covered in the order-tracking tutorial.

Contact-lens reorders and lens questions

Reorders look like simple repeat purchases, but in an optician they usually depend on up-to-date checks and a valid specification, and the rules that apply to you vary. So the chatbot's job is to collect, not to approve. A sensible flow:

1. Ask for the name, date of birth or customer reference your system uses, the lens brand and type the customer wears, and the delivery or collection choice.
2. Show a short statement from your fact sheet: "We confirm your lens details and check your aftercare is up to date before sending an order. A team member will confirm by [method] within [time]."
3. Create an enquiry for staff, flagged "reorder". Staff check the record and approve, ask for a check-up or decline.

Lens questions are different from lens orders. The bot can explain, in the practice's own words, the difference between lens types or coatings, or how long new varifocals take to get used to if your dispensers have written that down. It should stop when a customer asks "which lens is best for me?", and offer a dispensing appointment. The aim is to save repetitive explaining, not to replace the conversation that sells the right lens. If you are also considering virtual try-on and frame advice, that is a separate decision covered in AI frame advice and try-on for opticians.

The hand-off for eye symptoms

This is the part to get right before anything goes live. A chatbot that chats cheerfully with someone describing sudden vision loss is a serious failure, however good the rest is. The design is deliberately blunt:

• A list of trigger phrases and topics, written with your optometrists: pain, redness, sudden blurring or loss of vision, sudden flashes, a shower of new floaters, a shadow or curtain across the vision, light sensitivity, injury, chemical contact, a red or painful eye in a contact-lens wearer, and anything similar. The model is told to treat any mention of these as a hand-over, even in a message about something else.
• A fixed message your clinician approves. It says the practice cannot assess symptoms in a chat, gives the phone number and opening hours, says what to do outside hours according to your own policy, and tells the person to seek urgent care if their situation feels urgent. It does not diagnose, reassure or suggest waiting.
• A separate first line for chemical splashes and injuries, because these cannot wait for opening hours. Public first-aid advice is to rinse the eye with plenty of clean water straight away and get emergency help for a strong chemical or a penetrating injury. Your optometrist decides the exact wording the bot shows. For a contact-lens wearer with a sore or red eye, the message can also carry the standard advice to take the lenses out.
• No further chat in that conversation about the symptom. The bot offers to pass a message to a person and stops.

Test it with at least thirty realistic phrases, including vague ones ("my eye feels weird") and ones buried in other questions ("can I book a test, I keep seeing sparkles in one eye"). Record the failures, tighten the instructions and test again. Repeat after every change to the bot's settings.

160 enquiries a week at a two-site optician

Say a two-site independent optician gets about 160 enquiries a week by phone, email and web form. Staff log them for a fortnight and sort them. About 45 per cent are opening hours, test bookings, "what do I bring" and repair status. Roughly 30 per cent are bookings and reorders that need a staff check. About 15 per cent are lens, prescription or aftercare questions, and 10 per cent are complaints, odd queries and the occasional symptom.

The practice writes a two-page fact sheet, sets up a chatbot in request-only mode and publishes it on its website. In the first month, the bot handles most of the first group without staff, and passes the second group on as complete enquiries with the details already gathered. If 160 enquiries a week produce roughly 72 simple ones and the bot resolves most of them, staff might save a few hours a week, mostly in phone time. These figures are an illustration, not a forecast, and the practice measures its own.

The check is simple: each week a manager reads twenty chat transcripts, marks any wrong or poor answers and fixes the fact sheet or instructions. Any symptom mention that was not handed over is treated as an incident and fixed the same day.

A fact-sheet extract and the answers it produces

To see how the fact sheet drives the bot, here is a small filled-in extract and the replies it should produce. Everything in the left column is a placeholder for your own policy, written in the voice you want the bot to use.

The last row is the one to copy: where the fact sheet is silent, the bot says so and offers a person. Build that behaviour into the instructions and test it, because models are inclined to fill silences with plausible content.

The instruction set behind the bot

Whatever tool you choose, it will have a place for instructions. Write them plainly. An illustrative set:

You are the website assistant for [practice name], an optician. You are an AI, and you say so at the start of each chat.
Answer ONLY from the fact sheet provided. If the answer is not there, say you do not know and offer to pass the question to the team.
Never give medical, clinical or prescription advice. Never interpret symptoms or results.
If the customer mentions pain, redness, flashes, floaters, a sudden change in vision, light sensitivity, an injury, a chemical in the eye, or anything similar, stop and show the APPROVED SYMPTOM MESSAGE below, then offer to pass a message to the team. Do not continue discussing the symptom.
Never promise a time, a price or an outcome that is not in the fact sheet or in the status data.
Keep replies under 80 words, in plain British English. Always include the practice phone number and "talk to a person" option.
APPROVED SYMPTOM MESSAGE: [the clinician's exact wording]

Keep a copy of the instructions with a version number and date, and rerun your thirty symptom tests whenever you change a word. Small edits sometimes change behaviour in surprising ways.

Reading transcripts: what a good and a poor chat look like

The weekly transcript review is where the bot improves. A good chat is short: greeting, the question answered from the fact sheet, an offer of a booking, a clear route to a person. A poor chat usually shows one of three patterns. In the first, the bot answers a lens question with a confident paragraph that appears nowhere in the fact sheet; the fix is to add the practice's own answer or to make the bot decline. In the second, the customer asks the same thing three times in different words because the first reply missed the point; the fix is to rewrite the fact-sheet entry in the customer's words. In the third, the customer leaves after being asked for too many details before getting help; the fix is to answer first and ask for details only when a booking or reorder is requested.

Keep a simple log of each problem, the change you made and the date. After a month, the log is also your evidence of whether the bot deserves a bigger job.

Five test conversations to run before launch

• The late-night booker. "Can I book an eye test next Saturday morning?" The bot should offer the request flow or live slots, not promise a time it cannot confirm.
• The impatient repairer. "My glasses were meant to be ready yesterday." The bot should check status if it can, apologise without blame and offer a staff call-back if it cannot.
• The lens runner. "I have two days of lenses left, can you send more?" The bot should collect details and flag as urgent for staff, never promise dispatch.
• The symptom aside. "Booking for my son, also his eye has been red and sore all week." Hand-over message first. The bot should not slot him into a routine eye test; it passes the request to staff flagged as a symptom, so a clinician decides how soon he is seen.
• The price hunter. "How much are varifocals?" Answer only from the fact sheet. If there is a "from" price, quote it with the condition. If not, offer an appointment or a call.

Choosing a tool without over-buying

Many chatbot products let you upload a fact sheet and embed a widget on a website; some connect to booking systems and messaging channels. Prices and features change often, so compare current vendor pages rather than relying on a summary, and ask each vendor four questions: can the bot be restricted to your uploaded content, can I see and export every transcript, can I set a fixed hand-over message that overrides the model, and what happens to the data. Start with the cheapest plan that meets those four. A simple set-up that works is better than a connected one that is hard to test. To see where a chatbot sits among the other jobs AI can take on in a practice, read practical AI uses for an independent optician.

Five ways an optician chatbot lets customers down

Over-friendly answers about eyes. The model drifts into general eye-care advice because it has read a lot of it. The instruction "answer only from the fact sheet" must be explicit, and transcripts must be reviewed.

Stale information. Opening hours change at Christmas, the fact sheet does not, and the bot repeats the old hours for a month. Give the fact sheet an owner and a monthly review date.

No human in sight. A customer cannot find how to reach staff. Put the phone number and a "talk to a person" button in every reply, not just the greeting.

Promising what staff cannot deliver. "Your lenses will arrive tomorrow." Only staff can promise that. Ban time promises unless they come from a staff-updated status.

Unrecorded hand-overs. A symptom is flagged and nobody reads the flag. Decide who checks the flagged enquiries, how often, and what they do. A hand-over with no owner is only a nicer-looking dead end. And because a chat can arrive at 11 pm, the symptom message itself must tell the person what to do now; the staff flag is follow-up, not the safety net.

Four numbers to track for the first two months

Track four numbers for the first two months: chats a week, the share resolved without staff, the share handed over, and the number of wrong answers found in the weekly transcript review. Add a simple "was this helpful?" button and read the unhappy ones. If the wrong-answer count is not falling towards zero, narrow the bot's job rather than widening it. A chatbot that reliably answers ten questions is more useful than one that unreliably answers a hundred. For the receptionist side of the same question, the booking-mistakes tutorial shows what errors look like and how to repair them.

Questions opticians ask about chatbots

Can a chatbot tell a customer whether they need an eye test?

It can share the practice's own guidance on how often tests are usually recommended, if you supply that wording. It should not judge an individual's eyes. Anyone mentioning symptoms, or unsure, is offered a booking or a call from staff.

Should a chatbot take contact-lens reorders?

It can collect the request and check basic details against your rules, but staff should confirm that the customer's specification and aftercare are in date before anything ships. A safe pattern is for the bot to gather the order and a team member to approve it.

What must the chatbot do if someone describes an eye problem?

Stop answering and show a fixed message you have approved, with how to reach a clinician or urgent care. It should not diagnose, reassure or suggest waiting. Test this with realistic phrases before launch.

Do we need to tell customers they are talking to an AI?

Yes, say so at the start of the chat and make it easy to reach a person. Check the rules that apply to your business as well, since disclosure requirements differ between places and are changing.

Can the chatbot see patient records?

For most practices it should not. Keep it to general information and booking. If you later connect it to your booking system, limit what it can read to availability and an enquiry record, and review that with your data-protection adviser.

Further reads

• AI for Optometrists: Contact Lens Aftercare and Reorder Messages — Aftercare and reorder messages that pair with the chatbot's reorder flow.
• AI Receptionist for Optometrists: Eye Test Bookings and Recalls — The phone side of eye test bookings and recalls.
• AI Receptionist or Chatbot: Which Suits a Cleaning Company? — Choosing between a phone assistant and a web chatbot.
• Should an AI Chatbot Answer Allergen Questions for Your Restaurant? — Another safety-sensitive question set with a staff hand-off.
• Can a Hotel Chatbot Take Bookings or Only Answer Questions? — Booking versus answering: where the bot's job ends.
• How Independent Opticians Can Use AI for Recalls and Bookings — Recall and booking routines for an independent practice.
• AI Receptionist for Opticians: Cost, Bookings and Frame Queries — Which optician calls an AI receptionist can take, which must reach a person, and what each plan costs at a realistic call volume.
• Optician Website: AI-Drafted Eye-Test Pages, Frame Ranges and FAQs — Prompts and a sign-off routine for drafting an optician's eye-test page, frame ranges and FAQs with AI, with booking wording and no clinical claims.
• Optometry Insurance Verification: A Front-Desk Workflow With AI — A front-desk routine for checking vision and medical eye insurance before the visit, with a call script, note template and safe uses of AI.
• AI in an Optometry Practice: A Week-by-Week First-Month Plan — A four-week plan for an optometry practice: phones, recalls, exam notes and marketing, with what to measure and what to leave out.
• GDPR in an Optical Practice: Which AI Tools May Touch Records — A scored 12-point checklist for vetting any AI tool before it touches eye-test records, with four tools worked through and the paperwork behind each.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: AAO EyeSmart page on torn or detached retina symptoms, nhs.uk eye injuries page (rinsing after chemical contact) and CDC contact-lens prevention page, all checked 4 November 2026. Practice policies in examples are placeholders for your own approved wording.