Independent opticians can use AI to clean the recall list, write personal reminders timed to each patient's due date and preferred channel, follow up twice, and answer booking questions or fill cancellations through a website or phone assistant. The optometrist sets each recall interval; AI handles wording, sorting and replies, and anything about symptoms goes to a person.
Most practices treat recalls as one job, "you're due," when they are really four: routine eye tests, contact lens aftercare, glasses ready to collect, and clinical follow-ups the optometrist asked for. They differ in urgency and wording, and the clinical ones matter most. A patient who ignores a routine reminder loses a little; a patient who ignores a requested six-month recheck may lose more.
Four recalls, four different messages
| Recall | Trigger | When to send | Tone | What AI does |
|---|---|---|---|---|
| Routine eye test | Recall date set by the optometrist | About 4 weeks before, then 2 follow-ups | Friendly, low pressure | Personal wording; reply sorting |
| Contact lens aftercare | Aftercare interval for the lens wearer | 3-4 weeks before; also when lens supplies run low | Practical: needed to keep supplying lenses | Ties the reminder to their lens supply |
| Glasses or lenses ready | Order status changes to ready | Same day, reminder after 7 days | Short and clear | Little: a template does this well |
| Clinical follow-up | Optometrist requests a recheck (for example, eye pressures) | Before the requested date, with a personal phone call if unanswered | Calm, clear that the optometrist asked | Drafts wording; flags non-responders for staff to call |
The last row is the one to protect. Automated messages help, but a clinical follow-up that gets no response after two messages should land on a staff member's call list, not drift. Most practice management systems can flag these; if yours can't, keep a simple sheet of requested follow-ups with an owner and a date. A filled-in extract (illustrative):
| Record no. | Requested by | Recheck due | Messages sent | Status | Owner |
|---|---|---|---|---|---|
| 40217 | Optometrist A | Early March | 2 | Booked 4 March | Desk |
| 38804 | Optometrist B | Mid March | 2 | No reply: called 11 March, left message | Practice manager |
| 41590 | Optometrist A | Late March | 1 | Replied asking why; passed to optometrist | Optometrist A |
| 29931 | Optometrist B | Early April | 0 | First message due 10 March | Desk |
Row 38804 is the one the sheet exists for. Two unanswered messages and a voicemail means a second call, then a letter, with each attempt noted, so the practice can show it tried if the patient turns up with a problem a year later.
Clean the recall list before you automate it
Automating a messy list sends more wrong messages, faster. Run these checks on an export before the first campaign:
- Duplicates: the same patient entered twice with slightly different names or an old address. Both records get reminders, and one may hold out-of-date clinical details.
- Deceased patients: a recall sent to someone who has died is distressing for the family. Make sure these are flagged in your system and excluded.
- Patients who've moved or switched practice: anyone not seen for five or more years, or whose messages bounced, goes into a separate "check" group.
- Children: messages go to the parent or guardian contact, worded for the parent.
- Consent and channel: who agreed to texts, who to email, who opted out.
An assistant is good at spotting likely duplicates in a list: same date of birth and similar surnames, or the same phone number on two records. Do this only in a tool your practice has approved for patient data, or on an export with names replaced by record numbers; the patient data confidentiality checklist for small practices lists what to check first. In an illustrative practice with 6,500 patient records, a first pass flagged 140 likely duplicates and about 600 patients not seen in over five years. Staff confirm each duplicate before merging; the assistant only finds candidates.
Reminder wording that gets bookings
A before-and-after for a routine eye test reminder:
Before:
Our records show you are due for an eye examination. Please contact the practice to arrange an appointment.
After:
Hi [first name], it's nearly two years since your last eye test with us, so you're due for your next one in October. It takes about 30 minutes; please bring your current glasses. You can book online here: [link], or reply with a day that suits and we'll find a time. Reply STOP to opt out.
The prompt that produces the personal version from your recall export:
Write a reminder text (under 320 characters) for each patient row.
Use: first name, recall type, month due, and one practical detail
for that recall type (from the list below). Friendly, plain, no
exclamation marks, no health warnings or claims.
For CHILD rows, address the parent and use the child's first name.
For CLINICAL FOLLOW-UP rows, say "the optometrist asked to see you
again around [month]" and nothing about why.
End with the booking link and "Reply STOP to opt out."
If any field is missing, write CHECK instead of a message.
Practical details:
Eye test: takes about 30 minutes; bring current glasses.
Contact lens aftercare: wear your lenses to the appointment; bring
your lens case; we need this check to keep supplying your lenses.Illustrative outputs for the other types. For a child: "Hi, it's time for [child's first name]'s yearly eye test, due in November. It's relaxed and takes about 20 minutes. Book here: [link]." For contact lens aftercare: "Hi [first name], your contact lens check is due next month. Please wear your lenses to the appointment and bring your case. We need this check to keep supplying your lenses. Book here: [link]." For a clinical follow-up: "Hi [first name], the optometrist asked to see you again around March for a follow-up check. Please book a time here: [link], or call us on [number]."
Notice what the clinical wording leaves out. It doesn't say "to recheck your eye pressure," because a text is not the place for clinical detail and the patient already heard it in the consulting room. If a patient replies asking why, a person answers.
The prompt can only protect that if the export does too. A realistic slip on a first batch: the recall export included the optometrist's reason column, and one draft came back as "the optometrist asked to see you again around March to recheck your eye pressure, as it was slightly raised". The instruction said "nothing about why", but the reason was sitting in the row and the model used it. Remove clinical columns from the export altogether, so the only fields the AI sees are first name, recall type, month due and channel. In the same batch, the CHECK rule did its job on nine rows with no recall month; those went back to the optometrist to set, instead of going out as "due in [month]".
Online booking, and an assistant for the questions
A reminder only works if booking is easy. Two pieces make that happen.
Clear appointment types in the booking system. Illustrative lengths: eye test 30 minutes, child's eye test 20, contact lens aftercare 20, contact lens fitting 45, collection and adjustment 10. When patients choose the wrong type, the diary breaks, so name them plainly and put a one-line description under each. What to check in an AI booking system for appointment businesses covers the features to look for.
An assistant that answers the questions that stop people booking. On the website chat or the phone line out of hours, it needs your appointment types, prices, opening hours, parking and accessibility details, what to bring, and a firm set of handover rules. An illustrative exchange:
Patient: do i need to bring anything for an eye test, and can my 6 year old come at the same time
Assistant: Please bring your current glasses and a list of any medicines you take. Yes, we can book your son straight after you: a child's eye test takes about 20 minutes. Would a Saturday morning suit? I can show you the free times.
Patient: yes but also my right eye has been seeing flashes since yesterday
Assistant: Thank you for telling me. Flashes need to be looked at by our team today, so I'm passing this to a member of staff now. If you can, please also call us on [number].
The second reply is the important one. The assistant doesn't assess the symptom or reassure; it follows the practice's protocol and hands over. Write that protocol down, put the trigger words in the assistant's instructions (flashes, floaters, sudden loss of vision, painful red eye, injury), and test it with a dozen phrasings before going live.
Write the test phrasings the way worried patients type, not the way a textbook lists symptoms. An illustrative set: "keep seeing sparkly lights in one eye", "loads of new black dots since this morning", "like a curtain coming down on one side", "a shadow in the corner of my vision", "eye's gone red and it really hurts", "got cleaning spray in my eye", "can't see properly out of my left eye since last night", "my glasses feel wrong all of a sudden", plus the same messages with typos and no punctuation. In an illustrative first run, the assistant handed over on eleven of twelve but offered a routine eye test to "a shadow in the corner of my vision", because "shadow" wasn't in its list. Adding "shadow", "curtain" and "can't see" to the triggers, and telling it to hand over whenever a message describes a change in sight it isn't sure about, fixed it on the rerun. Run the whole set again after any change to the instructions.
Filling cancellations and cutting no-shows
Every empty test room slot costs the fee and any dispensing that would have followed. Two simple automations help:
- A confirmation message 48 hours before, with "Reply C to confirm or R to rearrange." Replies of R go to the desk or a rebooking link.
- A waiting list for short-notice slots. When a slot frees up, a message goes to patients who asked for an earlier appointment: "We've had a cancellation tomorrow at 11:30. Reply YES if you'd like it; first reply gets it." AI can draft these and sort the replies; the diary is updated by staff or the booking system.
Plan for "first reply gets it" producing several replies. Three patients answer YES within four minutes, and another replies after lunch, long after the slot has gone. Everyone after the first needs a prompt, kind answer, or the waiting list starts to feel like a lottery. Write it once and have the assistant send it word for word: "Sorry, that slot has just been taken. You're still on our list, and we'll message you as soon as another one comes up." Also decide how long a slot stays open. Offering tomorrow's 11:30 at 9pm and holding it until 10am works; offering it at 10:45 the same morning mostly produces apologies.
Reducing no-shows with AI reminders and automatic rebooking covers timing and wording in more depth, and the same recall logic in a neighbouring profession is set out in how dental practices use AI for recalls and reminders.
If you send these by WhatsApp rather than text, check costs: Meta's WhatsApp Business Platform pricing changes on 1 October 2026, making service replies within the 24-hour window chargeable after the first 1,000 a month per business number, with rates that vary by market.
An illustrative practice with 6,500 patients
Imagine an independent practice with two test rooms and about 6,500 patients on file. After cleaning, around 250 patients fall due for a routine test each month. Suppose postcards and generic texts currently bring about 35% of them in within eight weeks, roughly 88 tests.
After switching to personal reminders, two follow-ups and an online booking link, say that rises to 45%, about 112 tests, or 24 more a month. If an eye test brings in an illustrative $60 fee and 40% of patients go on to buy glasses or lenses averaging $250, each test visit is worth about $160 on average, so 24 extra tests are worth around $3,800 a month. Your own fees, dispensing rate and baseline will differ; the point is to measure the change against a clear starting figure.
The costs are small by comparison: text messages at anywhere from under a cent to several cents each depending on provider and country, an assistant plan for the staff member who reviews replies, and a website or phone assistant if you add one (check pricing, as these vary widely). Setup, mainly cleaning the list and writing the four message sets, takes two to three days spread over a month. For where recalls sit among other jobs worth handing to AI, see eight practical AI uses for an independent optician.
Measuring recall performance month by month
- Recall conversion by type: the share of each recall type booked within eight weeks of the first message.
- Clinical follow-ups outstanding: every requested recheck with no booking after two messages, and whether staff have called.
- Days from first message to booking: if most book after the second follow-up, the first message may be going too early.
- No-show rate before and after confirmations.
- Opt-outs and "wrong person" replies per 100 messages, which tell you whether the wording or the data needs fixing.
A month's figures read together tell you where to act (illustrative). Routine tests: 44% booked within eight weeks. Contact lens aftercare: 61%, helped by the lens-supply line. Clinical follow-ups: seven outstanding after two messages, five called, two not yet. Median days to booking: 12, with most bookings after the first follow-up. Opt-outs: 1 per 100 messages. "Wrong person" replies: 3 per 100. The two uncalled clinical follow-ups are the first job, whatever else the numbers say. The "wrong person" rate is the next: three in a hundred means changed mobile numbers, so ask for a number check at every visit and flag any record whose last text bounced. The median of 12 days with most booking after the follow-up suggests the first message could go a week later.
Once patients are booked, what happens in the chair matters too; explaining lens options and eye test results with AI help covers the conversation after the test.
Recall and booking questions from independent practices
Our practice management system already sends recalls. Why add AI?
Keep the system for timing and sending; it knows each patient's due date. AI adds value around it: cleaning the list before a run, writing better wording for each recall type, reading and sorting replies, and answering booking questions out of hours. If your system's reminders already book well, start with reply handling instead.
Can an AI assistant book appointments directly into our diary?
Only if your booking system offers online booking or an integration the assistant can use; many do. Start by letting it offer times and send the patient a booking link, with staff confirming anything unusual. Direct booking suits routine eye tests; contact lens fits and clinical follow-ups are better confirmed by a person.
What should happen when a patient mentions symptoms in a reply?
The assistant stops and passes the message to a person immediately, with the patient's contact details. Your practice should have a written protocol for urgent eye symptoms, such as sudden vision loss, flashes, new floaters or a painful red eye, and the assistant's only job is to follow it, never to assess the symptom itself.
Further reads
- How a Small Clinic Can Use AI at the Front Desk Safely — Using AI safely at a small clinic's front desk.
- Winning Back Lapsed Dental Patients With AI Messages — Messages for patients who haven't been in for years.
- Is AI Frame Advice and Virtual Try-On Worth It for Opticians? — Whether AI frame advice is worth adding for patients.
- AI Appointment Scheduling for Small Businesses: Tools and Setup — Compare appointment scheduling tools and their setup.
- Can AI Answer My Business Phone? What It Can and Can't Handle — What AI phone answering can and can't handle.
- How to Measure Whether Your AI Chatbot Is Actually Working — Judge whether a booking assistant is working.
- Can AI Handle Vaccination and Check-Up Reminders for a Vet? — What AI can and can't do for a vet practice's vaccination and check-up reminders, with a message sequence, a reply-triage prompt and worked numbers.
- 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: Meta WhatsApp Business Platform pricing update (effective 1 October 2026),.