How can a UK clinic create its ads with AI without breaking the rules?

Quick answer: Let AI draft the words, the pictures and the variations, and keep a person, with the clinician responsible, between every draft and the Publish button, because the UK advertising rules apply to an AI-made ad exactly as they apply to any other, and the clinic answers for it. The same three traps catch AI-made clinic ads as hand-made ones: naming or hinting at a prescription-only medicine such as Botox, which no paid ad may do; a before and after that is not a real, unretouched patient, which an AI image never is; and reaching under-18s. Below are the tools that help with each part of an ad, the rules quoted from their source, prompts that build the rules in, and a checklist to run before anything goes live.

This page is not legal advice. Every rule links to the regulator's own words, each source opened on 21 October 2026, and your own adviser decides what your clinic publishes.

Which AI tools help a clinic make its ads?

Three kinds, each good at one part of an ad: a general-purpose AI assistant for the words, image and video generators for the pictures, and the AI built into the ad platforms for variations and audiences.

The last row is the one most often missed. Google's ad tools ask for the address the ad will send people to, and "Google AI will use this page to summarise your products or services and to suggest and generate assets" (Google Ads Help). Google adds that generated assets "aren't guaranteed to be approved" and that you should review them "to ensure that they're accurate, not misleading and not in violation of any Google advertising policies or applicable laws before publishing them". If the page an ad points to names a prescription-only medicine, the tool has that word to borrow. Our reading: send ads to the consultation page, never to a treatment information page or a price list.

Why can't an AI-made ad mention Botox?

Because no ad may, whoever or whatever wrote it. Botox is a prescription-only medicine, and the law says "A person may not publish an advertisement that is likely to lead to the use of a prescription only medicine" (Human Medicines Regulations 2012, regulation 284). The advertising code's rule 12.12 says the same: "Prescription-only medicines or prescription-only medical treatments may not be advertised to the public" (CAP Code, section 12). The MHRA puts it in one line: "You can't advertise prescription-only medicines (POMs) to the general public" (GOV.UK, last updated 11 April 2025).

For ads in particular there is no exception at all. A clinic's website may mention Botox in narrow circumstances behind its consultation pages, but the CAP says "No reference to a POM should be made in a sponsored ad" (CAP advice on botulinum toxin, 29 October 2025), and its guide for practitioners is blunter: "You are not permitted to mention Botox in any paid advertising or on your own social media channels" (CAP Bitesize).

This is where AI needs watching. Tell a model not to write "Botox" and it may reach for another way of saying it, and the indirect ways are banned too:

• Other names for the same thing. Vistabel, Dysport, Bocouture and Azzalure are other brands of botulinum toxin (CAP advice).
• Descriptions that point to it. The phrases "wrinkle-relaxing treatments", "#brotox" and "#beautox" are named as indirect references (CAP Bitesize), and "Adding #botox (or similar, like #btx) breaks the rules" (ASA, 22 October 2020).
• Neutral-sounding labels that only fit it. In 2024 the ASA ruled that a clinic's Facebook ad for "cosmetic injections" by the number of areas advertised a prescription-only medicine, because the offer applied only to its anti-wrinkle treatment (ASA ruling, Valterous Ltd t/a Therapie Clinic, 18 December 2024).
• Other prescription-only medicines. The ASA names "weight loss injections and vitamin injections" alongside Botox among the prescription-only treatments it frequently deals with (ASA, 9 November 2023).

The regulator is watching, and now with AI. The ASA started using social media listening tools in 2020 to search Instagram for posts promoting prescription-only medicines, and reported that "More than 25,000 posts were taken down as a result of our work in 2022". It has since built a new set of tools on large language models to find them faster (ASA, 9 November 2023).

What an ad can promote is the consultation. The CAP's guide says you can advertise consultations for wrinkles on social media, in leaflets and on your website, and can promote "your clinic and other non-POM services, such as dermal fillers" and "your experience and qualifications" (CAP Bitesize).

Can AI make the before and after pictures for a clinic ad?

No. A before and after is offered as evidence of a real result, and an AI picture is not of a real patient, so it cannot be the proof the rules ask for. For a prescription-only treatment, no before and after may be used at all, real or not.

The CAP and the ASA "regard the use of 'before and after' photos in the same way as testimonials", and say advertisers "should hold signed and dated proof that the photos are genuine and have not been manipulated" (CAP advice on before and after photos, 5 June 2025). The same page records a ruling that reads like a warning about AI: in 2018 a laser clinic's before and after images were "not of their own customers, but were indicative of the results one could achieve", and the complaint was upheld. A generated picture is "indicative" by construction.

Three related lines matter just as much:

• A prescription-only treatment. "'Before and after' imagery of a prescription-only product, even in isolation without any accompanying claims, is very likely to be seen as an implied ad for a prescription-only product" (CAP).
• AI editing of a real photo. The rule on retouching applies whatever tool does it: "don't enhance or retouch any areas of the photo related to the treatment", and keep a record of what the photo looked like before any editing (CAP Bitesize). An "enhance" button that smooths skin is a retouch.
• A label does not fix it. "Adding a caption or disclaimer like 'filter used' or 'digitally enhanced' probably won't be enough to fix a misleading impression" (CAP Bitesize). Our reading: "made with AI" under a misleading picture is the same kind of caption. The ASA's own advice on AI says: "If AI in your ad creates a misleading impression of your product, then it will likely hit a regulatory firewall" (ASA, 12 June 2025).

What AI can do with real, consented photos of a treatment that is not prescription-only: crop them to the same frame, lay them out, write the caption. It should not touch the face.

Can an AI-generated doctor or patient appear in a clinic ad?

Not as someone they are not. A generated face presented as your practitioner, or as a patient, tells the viewer something untrue about who treats them or what happened to someone, and a health professional, real or generated, may not endorse a medicine at all.

• Rule 12.18: "Marketers must not use health professionals or celebrities to endorse medicines" (CAP Code, section 12).
• On deepfakes: "If an AI-generated celebrity 'endorsement' looks real enough to mislead consumers, it's likely breaching the Code" (ASA, 11 June 2026).
• Testimonials: advertisers "must hold documentary evidence that a testimonial or endorsement used in a marketing communication is genuine" (rule 3.47), and "must not contain fake consumer reviews" (rule 3.44) (CAP Code, section 3).
• Who treats: "Be clear about who will be administering the treatment, their qualifications and where it will take place" (CAP Bitesize).

The platforms may say so for you. Meta is moving its "AI info" labels into "About this ad", a page "within the three-dot menu on every ad", and says it will "begin automatically detecting ads created or edited using third-party AI tools", so an ad made with another company's AI can be labelled too (Meta, updated 1 June 2026). The safer use of AI for people is the other way round: film your real practitioners and let AI help cut, subtitle and resize the footage.

How do you keep AI-targeted clinic ads away from under-18s?

Set 18 as the minimum age yourself, keep it fixed when a tool offers to find the audience for you, and keep clear of places where under-18s are a quarter or more of the audience. Rule 12.25 says ads for cosmetic interventions "must not be directed at those aged below 18 years through the selection of media or context in which they appear" (CAP Code, section 12), in force since 25 May 2022. It covers fillers and other non-surgical treatments, not only prescription ones.

The CAP's guide turns that into two working tests (CAP Bitesize):

• Online and in print, these ads "can't appear in media where under-18s make up 25% or more of the audience", and on social media you must take "all reasonable steps to avoid reaching under-18s".
• Working with a content creator, check whether under-18s make up 25% or more of their following.

Automated targeting does not move the responsibility. The ASA ruled against an ad made with Google's automated tools and stated that "the advertiser was still primarily responsible for ensuring compliance with the CAP Code" (ASA, 11 June 2026). Our reading: where an ad tool offers to widen your audience, check in its settings that the minimum age is a hard limit, not a suggestion.

What else must a clinic ad avoid, however it was made?

Pressure, trivialising, and prices that hide charges. These are the habits of general advertising copy, and an assistant asked for "an ad that converts" can bring them with it, so name them in the prompt and check for them after.

• Trivialising. The ASA has ruled against "20-minute procedure that can be performed on your lunch break" (CAP Bitesize).
• Time pressure. No "short response times, or phrases like 'Hurry, offer ends at midnight' or 'Only three days remaining'" (the same page).
• Exploiting insecurities. In 2023 the ASA ruled against "Back to School Botox", which it found "exploited women's insecurities about aging" as well as advertising a prescription-only medicine (CAP advice).
• Hidden charges. Prices must "include all non-optional charges", such as booking fees or deposits (CAP Bitesize).
• Offers on prescription treatments. "Promotional marketing is still advertising", and it is not allowed for Botox or other prescription-only medicines (the same page).

What prompts help AI write clinic ads within the rules?

Prompts that state the rules before the task: what the ad promotes, the words it may never use, the claims it may never make, and who the practitioner really is. Three that work as starting points, with the square brackets filled in by you:

For the words:

You are drafting a paid social ad for a UK aesthetic clinic, [clinic name], in [town].
The ad promotes a consultation for [concern, e.g. lines and wrinkles]. It does not promote any treatment or product.
Rules you must follow, without exception:
- Never name or hint at a prescription-only medicine. Banned words include Botox, botulinum, toxin, tox, Vistabel, Dysport, Bocouture, Azzalure, anti-wrinkle injections, wrinkle-relaxing, baby Botox, Brotox, Beautox, and any hashtag built from them.
- No promised or "guaranteed" result, no "risk free" or "pain free", no before and after claims.
- No deadlines, countdowns, limited slots, discounts or giveaways.
- Nothing that suggests the reader needs to change how they look.
- The consultation is given by [name], [profession and registration]. Do not describe anyone else.
Write five versions of the main text and five short headlines. After them, list any word you used that a reader could take as a reference to a medicine.

For a picture:

A calm, bright consultation room in a small UK clinic: two chairs angled towards each other, a desk, natural daylight, a plant. No people, no faces, no hands. No syringes, needles, vials, packaging or medical products. No text in the image.

For a second look at a draft:

Below is a draft ad for a UK aesthetic clinic. Check it against this list and quote every line that might break one of these rules: it names or implies a prescription-only medicine (including brand names, nicknames and hashtags); it promises or implies a result; it uses time pressure or an offer on a prescription treatment; it makes treatment sound trivial; it suggests the reader should be unhappy with how they look; it describes a person, title or qualification. Do not rewrite it. Only list the lines and the rule each might break.

[paste the draft]

The picture prompt leaves out needles and vials on our reading, not on a ruling: in a clinic ad they point at injectables, and the CAP names "depicting syringes" when it warns other products against implying results comparable to Botox (CAP advice). The third prompt is a second pair of eyes, not a sign-off. A model asked to find problems can miss one, and it cannot know what your clinic offers or who your practitioners are.

What should a clinic check before an AI-made ad goes live?

Every line below, by a person, on the ad as it will run, including the version the platform generated rather than the one you wrote.

Words

1. No prescription-only medicine is named, nicknamed, hashtagged or hinted at, in the text, the headline, the image or the comments you post.
2. The ad promotes a consultation, a non-prescription treatment or the clinic itself.
3. No promised result, no "risk free", nothing that makes treatment sound trivial.
4. No countdowns, "limited slots" or offers that rush a decision, and no offer of any kind on a prescription-only treatment.
5. Any price includes every non-optional charge, such as a booking fee or deposit.

Pictures and video

1. No before and after for a prescription-only treatment. Any other before and after is a real patient of the clinic, with signed, dated consent, unretouched in the treated area, with the original kept on file.
2. No generated face is presented as a patient, a practitioner or anyone who has had treatment.
3. No filter or "enhance" on any person in the ad.

People

1. Every practitioner named or shown is real, and every qualification mentioned is one you hold evidence for.
2. No health professional or celebrity is endorsing a medicine.

Audience and destination

1. The minimum age is 18, and any automated audience setting keeps it as a limit.
2. The ad links to the consultation page, and nothing reachable from the ad's first page names a prescription-only medicine.

Sign-off

1. The clinician responsible for the treatment has read the ad as it will run.

Who is responsible when an AI tool made the ad?

The clinic, and the professionals who treat there. The ASA says that "even though the ad was produced using Google's automated tools, the advertiser was still primarily responsible for ensuring compliance with the CAP Code" (ASA, 11 June 2026), and that "the CAP and BCAP Codes do not contain specific rules relating to AI at this time", so the existing rules simply apply (ASA, 12 June 2025). A nurse's own code adds a personal duty: anything published for their services must be "accurate, responsible, ethical, do not mislead or exploit vulnerabilities and accurately reflect your relevant skills, experience and qualifications" (NMC Code, 21.4).

What else do clinics ask about AI and their ads?

Do we have to say an ad was made with AI?

The CAP Code has no rule that says so; what it asks is that the ad does not mislead (ASA, 12 June 2025). Where AI made a person or a result look real, saying so will not rescue the ad, and the honest fix is to remove it. Some platforms add their own label to images their tools generated.

Can we promote our consultations on Instagram?

Yes. A consultation for lines and wrinkles may be advertised on social media, as long as the post, its hashtags and its picture say nothing about a prescription-only medicine (CAP Bitesize).

Where should the ad send people?

To a page built around the consultation, so that a visitor meets the consultation first, and an ad tool that takes its words from the page finds nothing it should not borrow. How a clinic website is arranged for that is in how to create a website for a UK aesthetic clinic.

Who can check an ad before it runs?

Your own adviser. The ASA's Copy Advice team also answers questions about specific ads and websites (CAP Bitesize).

Where do we start if nobody at the clinic uses AI yet?

With one job you can check at a glance, such as drafting five versions of a consultation post. How to start using AI in your business sets out the first steps and what to keep out of an AI tool, and practical AI training does it with you on your own work.

An ad is only as good as the page it lands on. How we build websites walks through the steps, from a first call to handover, and how this site was built shows the method in the open: AI does the making, and a person checks every piece before it ships. To see how your own site measures up first, ask for a free website audit for your clinic: every finding checked by a person, including against the advertising rules.