Patient Recall Letters for Reviews and Screenings: AI Drafts

To write patient recall letters for reviews and screenings with AI, draft one template per recall type with merge placeholders, keep the wording factual and low-pressure, add a clear opt-out and contact route, then have a person check every template before use. Check the template, not each letter, and never paste patient details into the prompt.

Recall is repetitive, high-volume writing, which suits AI, and it is also a place where small mistakes do harm: a letter to the wrong person, a hint of a condition on an envelope, a reminder sent to someone who asked not to be contacted. The method below separates the two jobs. AI writes the words once, a person approves them, and your own system merges the names and dates.

Follow me on Instagram@sagnikteaches

What a recall letter has to do

A recall letter does five things in a few short paragraphs: tells the patient who is writing, says that something is due, says what to do next, makes it easy to opt out or ask a question and avoids saying anything the patient would not want a stranger to read. The recall pages that rank for this search are mostly software vendors describing automated reminders. They say little about the wording itself, which is where most of the risk and most of the response rate lives.

Connect on LinkedInSagnik Bhattacharya

Before drafting, decide which recall types you send. Examples, adapted to your practice: an annual review, a long-term condition check, a routine screening invitation, a medicines review, a follow-up after treatment and a "we have not seen you for a while" letter. Each has a different tone and different content, so each gets its own template.

Subscribe on YouTube@sagnikteaches

The wording rules, in one list

Write these on one page and give it to whoever approves templates. They are the rules the AI is given, and the rules the checker uses.

1. Say who and why, briefly. "Your annual review is due" is enough. No diagnosis, no medicine names, no test results.
2. One action. Call, book online or reply. Do not offer three.
3. No fear and no guilt. Avoid "you must", "failure to attend" and "serious risk". State the benefit in neutral words.
4. Plain reading level. Short sentences, no abbreviations a patient would not use.
5. Opt-out in every letter. How to say they do not want reminders, and a promise that it will not affect their care, if that is true.
6. No identifiers beyond what is needed. The letter may be read by family members or colleagues.
7. Dates and numbers come from the system. The AI writes placeholders; the mail-merge fills them.
8. Do not promise. No guaranteed appointment times or results.

Rule 3 is the one the AI breaks most often. Left alone, it writes warm reassurance with an undercurrent of urgency: "It's really important that you don't miss this". That sentence might be accurate, but it is the practice's job to choose the strength of its language, not the model's.

A prompt for one template

Draft a recall letter template for a small health practice.
Recall type: annual review for a long-term health condition.
Do not name any condition or medicine. Use placeholders in square
brackets: [first name], [practice name], [review type], [booking link],
[phone number], [opening hours], [sender name].

Rules: plain English, about 120 words, friendly and calm. One action:
book by phone or online. No urgency, guilt or threats. Do not promise
appointment times. Include an opt-out line: "If you would rather we
did not send you reminders like this, let us know and we will stop."
End with "Please ignore this letter if you have already booked."

Give two versions: a letter and a text message under 300 characters.

That prompt names no patient and no condition. The "annual review for a long-term condition" phrasing is generic on purpose. A typical result (illustrative):

Dear [first name],
Your [review type] at [practice name] is due. These reviews give us a chance to see how things are going and answer any questions you have.
You can book online at [booking link] or call us on [phone number] ([opening hours]).
If you would rather we did not send you reminders like this, let us know and we will stop.
Please ignore this letter if you have already booked.
Kind regards,
[sender name]

What I would fix: "to see how things are going" may feel vague if the patient has no idea what the review involves, so I would add a line about what to expect and how long it takes, but only if your practice can state it accurately. Check that the letter works when [review type] is replaced by each of your real values, such as "annual review", "health check" and "medication review". A sentence that reads well with one can be clumsy with another. Run the stranger test on the values themselves too: a value such as "asthma review" puts a condition on the page, which breaks rule 1 as surely as a sentence would, so keep the condition in your system and give the letter a generic label.

Templates by recall type

1. Annual or periodic review

The simplest letter. The only variables are the review name and the booking route. The example above is the basis. A second reminder should be shorter still, with an added line along the lines of "we wrote a few weeks ago", and nothing that implies the patient is at fault.

2. Screening invitation

Screening has more rules, because programmes are often run to a published protocol that sets eligibility, timing and wording. If your practice sends invitations on behalf of a programme, use the programme's own approved text as the source and let AI do only light edits for readability, then compare the result with the original line by line. If the practice sends its own reminders about a programme run elsewhere, say so plainly: "You may be due an invitation for [screening name]. If you have already had it, no need to reply." Do not add statistics or claims about effectiveness unless they come from the programme's own current material.

3. Medicines or long-term condition review

Be the most careful here, because the letter itself may disclose a condition. Use the generic term in the letter and move specifics to the appointment: "your regular review" rather than naming the condition. If the review is about a medicine, do not name it. Where the patient may have been told they must attend for safe prescribing, your clinician decides the wording, and AI should not touch it.

4. Post-treatment follow-up

This one asks how things are going and offers a check. A good template asks one open question and gives a route to book, and says what to do if they are worried, using your own approved safety-net wording. Never let the AI write the safety-net line, because it is clinical advice.

5. "We have not seen you for a while"

The tone is welcoming, not scolding. Something like: "We noticed it has been a while since your last visit, and we would be glad to see you." Avoid "you have missed". It pairs naturally with the approach in winning back lapsed patients, which covers the wider message sequence.

For dental practices, the mechanics of recall intervals and reminders are covered in how dental practices use AI for recalls and reminders, and AI for dental hygienists shows how recall messages can include a line of patient education.

The opt-out, built so it works

An opt-out line in the letter is only as good as the process behind it. Decide four things. First, how patients opt out: phone, a reply email, a tick on a form or a link. Second, who records it and where, so that the next mailing excludes them. Third, how fast it takes effect, ideally the same working day. Fourth, what it covers: a patient may want no text reminders but still welcome letters, or may want no marketing but still need clinical recalls. Your privacy rules and communication preferences should reflect that distinction, so check with whoever advises on your data obligations which messages count as marketing in your setting.

Test the process with a dummy patient each quarter: mark them as opted out, run the recall list and confirm they do not appear. A letter that goes to someone who opted out is the single most damaging mistake in this whole workflow, so it is worth ten minutes of checking.

Before and after: rewriting a harsh letter

Here is an illustrative old letter of the kind many practices still send: "You have failed to attend your review. It is essential that you book immediately, as missing this appointment could put your health at risk. Contact us urgently." Run it past the rules. It blames the patient, creates alarm, hints at a health risk and demands immediate action. A rewrite that passes: "Dear [first name], we noticed your [review type] is still due. We would be glad to see you, and you can book online at [booking link] or call [phone number]. If you would rather we did not send reminders like this, let us know and we will stop." It is shorter, kinder and says the same thing that matters: a review is due and here is how to book. Whether to keep any stronger wording is a decision for a clinician, who may need a firmer line for a specific safety-related recall.

A filled-in screening reminder and a last letter

For a practice that reminds patients about a screening programme run elsewhere, an illustrative first reminder might read: "Dear [first name], you may be due an invitation for [screening name]. If you have already had your screening, or have an appointment, you can ignore this message. If you are unsure, call us on [phone number] and we will check. If you would rather we did not send reminders like this, let us know." The final letter in a sequence is shorter again, and I would end it with a line that closes the loop without pressure: "This is our last reminder for now. You are welcome to get in touch whenever you are ready." Then the outcome is recorded as "no response after two reminders", and the patient moves to whatever your protocol says happens next, which is a clinical or practice decision, not an AI one.

Keep the final-letter wording the same across patients, so nobody is treated differently by accident. If a clinician wants a particular patient contacted by phone instead, record that on the tracking list as a manual step with the date, and do not send a standard letter on top of it.

The tracking list

A recall list is a spreadsheet or a report from your practice system. It does not need to be fancy, but it needs these columns. Keep patient names out of any copy you share outside the system; use the record reference.

Do the sums to see whether the process works. Say 200 patients are due for an annual review in a quarter. A first letter goes to all 200 and, suppose, 90 book. A reminder goes to the remaining 110, and perhaps 30 more book. That leaves 80, who get a phone call from reception, a manual step, and say 20 of those book. You have reached 140 of 200, or 70 per cent, and 60 are recorded as declined or no response. These figures are an illustration, not a benchmark. The useful part is the structure: every patient ends with a recorded outcome, so nobody is lost between letters.

Checking a batch before it goes

Templates are checked once, but batches are checked every time. This is a five-minute routine.

1. Confirm the template version in use matches the approved one.
2. Open three random merged letters and read them in full: names, dates, review type.
3. Confirm the opt-out list was applied: compare the batch count with the due list minus opt-outs.
4. Check the sender name, phone number and booking link in the merged output.
5. Record the batch in the tracking list with your initials.

The merge is where the harm happens. If a placeholder is left unfilled, a patient receives "Dear [first name]", which is embarrassing but harmless. If the review type column is wrong, someone is invited to a review they do not need, which is worse. Reading three real merged letters catches both.

A realistic mistake and how it showed up

Picture a practice that asks AI to "write a recall letter for our diabetic patients" and pastes the result into the system as the template. Every envelope now carries a letter that names a condition, and one goes to a patient who shares a home with relatives who did not know the diagnosis. Nobody meant harm. The fix is a generic template with the condition name removed, a rule that letters never name a condition and a re-read of every template by someone who asks "what would a stranger learn from this?". That question is the best single check I know for recall wording.

A second example: text messages

Text recalls are short, easy to ignore and easy to get wrong. Ask AI for a version under 300 characters, with the sender name first so the patient knows who is writing, and a way to reply STOP or call. An illustrative result: "[practice name]: your [review type] is due. Book at [booking link] or call [phone number]. Reply STOP to opt out of reminders." Keep it to what is required. Do not mention conditions, medicines or tests in a text, since a phone screen can be seen by anyone. If your texting service uses a different opt-out keyword, put that one in the template, and check it works.

Using AI to review your existing letters

If you already have recall letters, you can use AI as a reviewer, not a writer. Paste the template (with no patient data) and ask it to list any wording that implies blame, urgency, a diagnosis or a promise, and any missing opt-out. Then decide what to change. It is a quick way to find a "failure to attend" in a letter written years ago. It will sometimes flag things that are fine, and miss others, so treat the list as a prompt for your own judgement.

Where it fits in the wider practice

Recall letters are one strand of patient communication. The phone side, handling calls from patients who reply to a letter, benefits from clear call flows; virtual receptionist call flows show how to structure them. A similar reminder pattern for animals is in vaccination and check-up reminders for a vet, which is useful for comparing tone. For other practice-admin prompts, see ChatGPT prompts for healthcare practice managers.

If you want someone to help set the process up, that is what my AI implementation consultation covers, and you can book a call.

Finally, think about language and access. If a share of your patients read another language, or need large print or an easy-read version, draft those from the approved English template and have a fluent reader or the relevant service check them before use. A translation that changes the meaning of the opt-out line is a serious fault, so check that line first.

A review rhythm

Review the templates twice a year, and whenever a protocol changes. Look at the tracking list: which letter gets the strongest response, which recall type has the most "no response", and how many opt-outs came in. A rise in opt-outs after a wording change is a signal to revert. Keep each template with a version number and the approver's initials, and keep the old ones, so you can say which wording a patient received on a given date. That small habit is also what you will want if a patient ever queries a letter.

Questions about writing recall letters with AI

Can I put a patient's name and condition in the AI prompt?

Avoid it. Draft the letter with placeholders such as [first name] and [recall type], then merge real details inside your practice system or mail-merge tool. The AI never needs to see who the letter is for or why they are on a list.

How many reminders should a recall sequence have?

Many practices send a first letter, one reminder and a final letter, spaced a few weeks apart, then stop and record the outcome. Your own protocol or contract may set the number, so check it. More than three starts to feel like pressure.

What should a letter say about the reason for the recall?

Only what is needed and what the patient would expect to see on an envelope or screen. Keep the wording general, for example 'your annual review is due', and leave clinical detail for the appointment, since letters can be read by other people.

What if a patient asks not to be contacted?

Record it on their file at once and remove them from the recall list for that channel. Every letter should say how to opt out, and the tracking list should show the date a request was received and acted on.

Further reads

• Patient Reviews and AI Replies: Staying Within Confidentiality — Keep replies to patients within confidentiality.
• Patient Data and AI: A Confidentiality Checklist for Small Practices — Check confidentiality before health data enters a tool.
• Patient Intake Automation in a Small Practice: Forms, AI Summaries — Capture consent and preferences when patients first register.
• Which Dental Front Desk Tasks Can AI Take Over? — See which front desk tasks suit AI drafting.
• Explaining Dental Treatment Plans to Patients With AI — Explain treatment plans once a recalled patient attends.
• AI Marketing Automation for Doctors: Recall, Reminder and Review Flows — Build recall, reminder and review-request flows in the practice software you already own, with consent, opt-outs and a worked month of numbers.
• Orthodontic Payment Plan Template: Fee Scripts Drafted With AI — A fillable orthodontic payment plan layout, three parent fee scripts and the checks that stop AI drafts quoting the wrong numbers.
• Orthodontic Referral Form Template and an AI Completeness Check — A field-by-field orthodontic referral form, a radiograph and consent checklist and a rule-based AI check for missing details.
• Orthodontic Staff Training: AI Role-Play Scripts for Phone Calls — Role-play prompts, a call-type list, a scoring sheet and a quiz bank for orthodontic front-desk phone training, built with AI and reviewed by a coordinator.
• Patient Participation Group: Agendas, Minutes and Newsletters With AI — A practice manager's AI workflow for patient participation group agendas, minutes drafted from notes and newsletters, with the checks to keep it safe.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general recall, consent and communication-preference practice; check the rules on health-related messages that apply to your practice. No vendor pricing is quoted.