AI Marketing for Medical Practices Without Making Health Claims

Use AI for what a practice can say without promising outcomes: who you are, what services you provide, how to book, opening hours, how to refer and how you reply to feedback. Draft from a fact sheet, ban result and comparison words, have a clinician check every medical statement and never put patient details in a reply.

That leaves a lot of useful ground. A practice that explains its services plainly, keeps its local listing accurate, answers reviews with care and keeps referring clinicians informed will usually do better than one that tries to sound impressive. And it will stay on the right side of advertising rules, which generally ask for claims you can prove and testimonials that do not stand in for proof.

Follow me on Instagram@sagnikteaches

What a health claim is and why AI produces them

A health claim, for marketing purposes, is a statement that a treatment achieves something. "Relieves", "reduces", "prevents", "restores", "cures", "proven", "fast", "safe", "better than" are the giveaway words. Advertising regulators expect a claim to be backed by evidence you can produce, and their published guidance on testimonials makes the point twice over: a claim inside a testimonial is treated like any other claim in the ad, and a patient saying a treatment worked is not substantiation. Models trained on the open web have seen thousands of clinic pages full of such phrases, so they reproduce them without being asked. Ask for "a page about our diabetes clinic" and you will often get promises about controlling blood sugar.

Connect on LinkedInSagnik Bhattacharya

The cure is to describe the service rather than its effect. Compare "Our diabetes clinic helps patients take control of their blood sugar" with "Our diabetes clinic offers review appointments with a doctor and a nurse, including blood tests, foot checks and a medication review. Book at [link]." The second says who, what and how. It makes no promise, and a clinician can verify each word.

Subscribe on YouTube@sagnikteaches

Build a fact sheet and a banned list first

Create one page per service containing only things you can document: who provides it, what an appointment includes, how long it takes, what to bring, how to book or be referred, what it costs if you publish fees. Add a banned list that every prompt carries:

• Result words: cure, fix, relieve, reverse, eliminate, prevent, restore, improve (unless backed by a source you hold).
• Speed and certainty words: fast, instant, guaranteed, proven, safe, effective, painless.
• Comparisons: better than, unlike other practices, alternative to medication.
• Statistics and awards not in your records.
• Any patient detail, even anonymised, if the case could be recognised.

The statements that stay are facts about the service. If a clinician wants to include a statement about benefit, they supply the source and the exact wording, and it goes in the fact sheet marked "clinician-approved, source: [name], checked [date]". The same approach underlies a claims register for chiropractors, which is worth a look if you want a ready-made layout.

Service pages that describe rather than promise

The page prompt

Write a service page for a medical practice. Use ONLY the facts below.
Do not use any word on the banned list. Do not describe what the treatment
achieves. Structure: one-line description; who provides the service; what an
appointment includes; how to book or be referred; what to bring; a short
note that a doctor will discuss options at the appointment. 250-350 words.
Plain English, reading age about 12. No exclamation marks.

BANNED: cure, fix, relieve, reverse, eliminate, prevent, restore, improve,
fast, instant, guaranteed, proven, safe, effective, painless, better than.

FACTS:
[paste the service fact sheet]

An illustrative page for a minor-skin-procedures service starts: "Minor skin procedures at [practice]. A doctor can assess a skin lesion and, where suitable, remove it in the practice. Appointments last [length]. Please bring a list of your medicines. To book, call [phone] or use [link]. After the appointment the doctor will explain what they found and the options." There is nothing to remove. What you check is that "where suitable" is accurate and that the page matches how the practice works.

The page audit

Run the audit prompt on every page, including ones written by people. It is often the existing pages, written years ago, that carry the worst claims.

Read the page below as an advertising reviewer for a medical practice.
List each sentence that states or implies a treatment result, speed,
safety, comparison, statistic or testimonial. Quote it and say which. If
none, write "No flags". Do not rewrite.

A realistic result on an old page: "Our laser clinic offers safe, effective treatment of veins with immediate results. Dr [name] is the region's leading specialist." Flags: "safe", "effective", "immediate results", "leading specialist". The rewrite keeps the facts: "The practice offers consultations about vein problems with Dr [name], who has [qualification, registration]. During the consultation the doctor will examine you, discuss options and explain risks." Replace "leading specialist" with the actual training and registration, which are checkable.

Review replies without patient detail

Google's policy on reviews prohibits incentives for reviews and selective solicitation, and you must treat anything a reviewer writes about their care as sensitive. A public reply that confirms the person was a patient, or names a condition, appointment or clinician, can itself disclose health information. Keep replies short and generic and have a person approve each one.

Draft a public reply to this review for a medical practice. Under 45 words.
Do not confirm the reviewer is a patient. Do not mention any condition,
treatment, appointment, date, clinician or fee. Do not defend our clinical
decisions. No offers. If negative, invite them to contact the practice
manager on [number]. Review: [text]

Three illustrative cases. A positive review that names a doctor: "Thank you for your kind feedback. We will share it with the team." A complaint about reception: "Thank you for telling us. We are sorry your experience was not what we aim for. Please contact our practice manager on [number] so we can look into it." A review that describes symptoms and a misdiagnosis: do not engage with the details. Use "We take concerns about care seriously and would like to discuss this with you directly. Please contact the practice manager on [number]," and tell the clinical lead and your indemnity provider, since this may become a complaint. The tutorial on patient reviews and confidentiality has more examples, including tone for positive and negative reviews.

Asking for reviews follows the same fence: neutral wording, no star request, no reward, and exclusions for sensitive visits and open complaints. The tutorial on recall, reminder and review flows covers the flow itself; here the rule is that the review request is not a marketing campaign and no one is nudged to leave a particular rating.

Referrer updates: marketing to other clinicians

If your practice receives referrals from other doctors, a short regular update is the most useful marketing you can do, and it needs no health claims. Referring clinicians want to know what you accept, how fast you respond, how to refer and what you will send back. A quarterly update of 200 words does that.

A filled-in referrer update

Subject: Update from [practice]: referral changes this quarter

Dear colleague,

Three practical updates from [practice].

1. Referral form: the new form is at [link]. It now asks for current
   medicines and the reason for urgency, which saves us a call.
2. Appointment availability: we have added [number] clinic sessions on
   [days]. Urgent referrals are triaged by a clinician within [timeframe].
3. Reports: we send a letter to the referrer after the first appointment
   and again when care is complete, by [secure method].

We do not offer any benefit for referrals. If you would like to talk
through a case, call [number].

[Name, role]

AI can tighten the draft and adapt it for different specialties from the same facts, but the timeframes and methods must be ones you actually meet. If you say "triaged within two working days" and then take five, you have broken a promise to the people whose trust you need. Never offer anything of value for referrals, and ask a qualified adviser before any hospitality or gift to a referring clinician, since rules on inducements are strict in many places. Do not name any patient in an update, even in a "case we helped with" story. The tutorial on referrer updates in a physiotherapy clinic shows the same pattern in another setting.

Patient education pages: draft with AI, verify like a clinician

A patient education page ("What to expect from a blood pressure review") is useful content, and it is also where medical errors enter. A model will write fluent advice that may be out of date, too general or wrong for your protocols. Use this order: the clinician supplies the key points and a source, AI turns them into plain English, a second clinician checks the result against the source, and the page carries a review date.

Turn the clinician's notes below into a patient information page of about
300 words in plain English (reading age about 12). Use ONLY the notes. Do
not add medical facts, numbers, doses or timescales that are not in them.
Add a closing line telling patients to contact the practice or, in an
emergency, emergency services. Mark any sentence you are unsure about with
[CHECK].

NOTES: [clinician notes and the source they came from]

Illustrative notes: "Review every 12 months; includes weight, blood pressure, blood test, medication check; takes 20 min; bring a list of medicines." The draft page follows those facts and no more. The usual failure is an extra helpful sentence such as "most people see improved readings within weeks", which the clinician did not write. The second clinician's job is to find sentences with no source and delete them. The tutorial on fact-checking patient education materials goes through the check in more detail.

The team and about pages

People choose a doctor partly on who they are, so team pages matter, and they are full of hidden claims. "Leading expert in", "specialist in the treatment of" and "years of experience helping patients overcome" all say more than the facts. The safe structure for each person: name and role, qualifications and registration, areas of clinical interest as stated by that person, languages spoken, which clinics they work in. A before-and-after, illustrative. Draft: "Dr [name] is a highly respected GP who has transformed the lives of hundreds of patients with chronic pain." Checked rewrite: "Dr [name] is a GP at [practice] with a clinical interest in long-term pain management. Qualified in [year]; registered with [register name]. Works on [days]." The model cannot know the number of patients, nor whether they were transformed, and neither can an advertiser prove it.

A Google Business Profile services list, filled in

Each service entry has a short description, so write each from its fact sheet row. Illustrative examples for a general practice: "Travel health: advice appointments before you travel, with a nurse. Please book at least [number] weeks ahead." "Minor surgery: assessment and removal of suitable skin lesions by a doctor. Referral not needed." "Blood tests: appointments with a healthcare assistant; fasting instructions are given when you book." Notice they are descriptive and each carries a practical instruction. Use the same sentences on the website so the two never disagree. Gemini can now edit some Business Profile fields, such as hours, the business description, attributes and links, but only for someone who manages a single verified profile from a personal Google account, not a work account. The services list is not among the fields Google says it edits, so enter these by hand.

A quick comparison: three practice types, three emphases

The risk column matters because it tells you where to put the strictest check. A specialist clinic can write accurate, helpful pages in a day, but each sentence about a procedure should be treated as presumptively a claim until a clinician has confirmed the source.

Local search: accuracy beats cleverness

When a patient searches for a clinic, Google shows a profile with the practice's name, address, hours, services and reviews. AI can help you fill it consistently, but the work is mostly accuracy.

1. Name, address and phone identical everywhere: the website, the profile, directories and referral forms. A variation such as "St" and "Street" is small but multiplies. Make one master record and copy from it.
2. Pick the primary category carefully and add the secondary categories that really describe services. Do not choose a category for a service you do not provide.
3. Services list: one line each, from the fact sheet, no result words. AI can draft the descriptions from the facts.
4. Hours: including public holiday changes, entered as special hours with dates.
5. Photos: real ones of the entrance, waiting area and team, with consent for any person. Avoid generated pictures of treatments.
6. Posts: short notices such as new clinic sessions or holiday closures. Events need start and end dates, and Google archives a post older than six months unless it has a date range.
7. Website pages: one page per service with the fact sheet content, and a "how to refer" page for clinicians.

The Google Business Profile performance report shows profile views, searches, calls, direction requests, website clicks and bookings. The searches figures update at the start of each month, so judge over three months rather than weeks. The report cannot tell you which searches became appointments, so add one field to your booking form or reception script, "How did you hear about us?", and log the answer in a fixed list (search, map, referral, friend, other) rather than free text you cannot count.

Newsletters to existing patients

A practice newsletter works if it is useful and sparing: two or three a year is plenty. Good subjects are service changes, holiday opening hours, how to use the online booking system, seasonal vaccination clinics (as logistics, not as advice), and a note from the practice manager. Contacting patients with marketing messages needs the consent rules described in the tutorial on recall, reminder and review flows, and anything clinical should come from a clinician. A realistic mistake: AI drafts a seasonal piece saying "flu jabs protect you and your family" and a patient asks the practice why it said "protect" when its own leaflet says "reduce the risk". Use the exact wording from the source the clinician provides, and attach the source line to the draft before it is approved.

Who checks what

Split the review so that nobody marks their own homework. The coordinator who drafts with AI checks style, facts about opening hours and links. A clinician checks every sentence that describes care, a treatment, a condition or a risk. The practice manager checks anything that touches complaints, fees or staff. Put the three initials and dates on a simple log sheet kept with the page. If a regulator, patient or competitor ever challenges a page, that log is how you show a process rather than a hope.

Keep the habit of dating the log, even for tiny edits such as a phone number. A page that was approved in March and altered in June without a check is the page that surprises you.

A worked quarter for a five-doctor practice

Illustrative figures for a five-doctor practice with a part-time marketing and reception coordinator. Month one: write 12 service fact sheets (6 hours), draft and audit the 12 pages (8 hours), correct and publish them (4 hours), clean up the profile (2 hours). That is 20 hours once. Month two: reply to every new review within three working days (about 1.5 hours a month), send the first referrer update (2 hours) and post two profile notices (1 hour). Month three: audit the older pages that were not rewritten (4 hours) and review the performance report (1 hour). Ongoing, the routine takes roughly five hours a month. At $25 an hour for the coordinator's time that is about $125 a month and 20 hours at $25, or $500, once. Whether that is worthwhile depends on bookings. Record the baseline: calls and booking clicks per month from the profile, and how many new patients said they found you online. Compare after three months and treat any change cautiously, since seasons and referrals vary.

Mistakes that show up in the audit

• The cloned page: AI writes ten service pages with the same structure and the same claims in a different order. Ask for varied page structure from the facts you give, and cut sentences that appear on more than one page.
• The invented credential: a draft adds "board-certified" or "award-winning" because it sounds right. Every credential comes from the registration record, nothing from the model.
• The stale page: a service that closed last year is still listed and still gets calls. Put a review date on each page and a calendar reminder.
• The over-eager chatbot: a website assistant answers a symptom question. Restrict it to an approved fact sheet, as in scoping a doctor's website chatbot.

Check the whole routine each quarter

1. Run the audit prompt on every live page and every profile post from the last three months.
2. Re-read the current advertising guidance that applies to you and compare it to your banned list.
3. Read ten review replies for names, conditions, dates and promises.
4. Check that opening hours, services and phone numbers match across the website, profile and referral forms.
5. Ask a clinician who did not write the content to read three pages cold and mark anything they could not defend.

None of this requires special software. It does require a named owner and a calendar reminder. If you would like help turning it into a routine for your practice, that is what my AI implementation consultation covers, and you can get in touch here.

Medical practice marketing and AI: common questions

What counts as a health claim on a practice website?

Any statement that a treatment achieves, prevents or relieves something, how fast, how often, or better than an alternative. Describing a service, who provides it and how to book is not a claim. Claims need evidence you hold, and testimonials do not replace it.

Can AI write patient education pages?

It can draft them, but a clinician must check every medical statement against a source you trust, since models state errors fluently. Keep pages general, link to a reputable public source where appropriate, and add a review date so stale advice does not sit on the site.

Is it acceptable to thank referrers with a gift?

Be careful. Rules on rewarding people for referrals are strict in many places. Share information, updates and clinic news freely, but offer nothing of value in return for referrals, and ask a qualified adviser before any hospitality or gift to a referring clinician.

How do I know the local search work is paying off?

Track calls, direction requests and booking-link clicks from your Google Business Profile report, and ask new patients how they found you at booking. The searches figures update at the start of each month, so judge over three months rather than a week.

Further reads

• Patient Satisfaction Survey With AI: Questions and Comment Summaries — Gather private feedback alongside public reviews.
• AI Assistant for a Medical Practice: Build a Staff Policy Q&A — Give staff a policy assistant that answers from your documents.
• GP Surgery Website: Plain-English Pages and Wording With AI — Plain-English website pages for a general practice.
• AI Tools for Medical Practices: A Stack for Three Doctors and Four Staff — A worked tool stack for a small medical practice.
• AI Website for Doctors: Page Plan, Contact Forms and Privacy Checks — Website pages and forms that respect privacy.
• AI for Healthcare Clinics: Referral Tracking and Chase Letters — Track referrals and chase missing letters with AI help.
• AI Search Tools for Medical Providers: How to Check the Citations — Check whether AI search tools cite your practice.
• Health Food Store Marketing Ideas: AI Posts Without Health Claims — A fact-sheet brief, a five-band claim check and eight post types that need no health claim, so a health food shop can use AI without risky wording.
• AI Secretary for Doctors: Referral Letters, Inbox, Insurer Forms — Set up three secretary jobs for a solo doctor: referral letters, inbox triage and insurer forms, with templates, a review step and a no-paste list.
• AI Tools for Dentists With No IT Team: Six to Try in the First Month — Six setup-light AI tools for a dental practice with no IT support, one a week, each with a data-agreement check and a first task to try.
• AI Tools for Dietitians: Admin, Content and Follow-Up by Job — One AI tool per job for a private-practice dietitian: admin and notes, content, and between-session follow-up, with costs, privacy checks and what to skip.
• Assisted Living Marketing Ideas: Eight Campaigns to Draft With AI — Eight practical campaigns for family decision-makers, each with a message angle, delivery channel, timing and human claims check.
• Home Care Agency Marketing Plan: Build One With AI and Local Data — A one-page plan built from your own local data: search terms, referral partners, family messages and a budget, with an AI prompt and a reality check.
• Prior Authorisation With AI: Drafting Letters and Tracking Status — A payer checklist, a redaction habit, a letter prompt and a status tracker speed up prior authorisation without handing patient data to a chatbot.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: published advertising-regulator guidance on health claims, testimonials and health-product substantiation; Google Maps user-generated content policy; Google Business Profile help on posts, performance and Gemini. Check the guidance that covers your own market.