AI-generated content in healthcare marketing

Writing tools that generate text, images, voices and video are now built into the software many practices already use: the website builder, the email platform, the design app, the ad account. A marketing coordinator can produce a blog post, a set of social images and a short patient story video in an afternoon. That is useful. It also makes it easy to publish something that would be a problem no matter who, or what, made it.

The good news for practice owners is that there is no separate rulebook to learn. The rules on deceptive advertising, testimonials, reviews, patient privacy and ad platform policy already apply to AI output exactly as they apply to anything else. This guide walks through where those rules bite hardest for AI-made marketing, and gives a review process that fits a small practice. How Google AI Overviews and AI Mode pick the pages they link to is covered in our AI Overviews guide for practices. Stock photos and paid actors raise the same questions, covered in our guide to stock photos, models and actors in healthcare ads. For how AI assistants find and cite practice pages, see our AI search visibility guide.

This is general information, not legal advice. Sources cited were checked on October 29, 2026: the FTC announcement of Operation AI Comply (September 25, 2024); the FTC announcement of the final rule on fake reviews and testimonials (August 14, 2024) and 16 CFR Part 465; the FTC Endorsement Guides, 16 CFR Part 255; the FTC order of December 22, 2025 setting aside a 2024 consent order against an AI review-writing service; the FTC Health Products Compliance Guidance (December 2022); HIPAA rules at 45 CFR 164.502, 164.508 and 164.514; Google Search Central guidance on generative AI content (last updated October 1, 2026) and spam policies (last updated August 28, 2026); and the Google Ads Misrepresentation and Manipulated media policies. Rules and platform policies change, so check the current version and ask counsel about your own materials.

There is no AI exemption

When the FTC announced its Operation AI Comply sweep in September 2024, the agency’s then chair said that using AI tools to trick, mislead or defraud people is illegal, and that there is no AI exemption from the laws on the books. Leadership at the FTC has changed since then, and one of the orders from that sweep was later set aside (more on that below). The underlying point has not changed: Section 5 of the FTC Act prohibits deceptive advertising, and the law looks at what the ad says to a reader, not at how the ad was produced.

For a practice, that means three things:

You are the advertiser. A claim on your website, in your ad or in your email is your claim, whether a staff member, an agency or a writing tool drafted it.
Health claims still need evidence. The FTC’s Health Products Compliance Guidance says health-related claims need competent and reliable scientific evidence. A tool that writes “clinically proven” or “lasting relief” into a draft has not supplied that evidence.
The net impression counts. The FTC judges the overall message a reasonable reader takes away, including what images and video imply. A generated photo can make a claim as clearly as a sentence can.

AI-written blog posts and service pages

Google said in February 2023 that appropriate use of AI or automation is not against its guidelines. Its current guidance on generative AI content, last updated October 1, 2026, adds the caveats that matter for a practice website. Generative models do not retrieve facts; they predict a likely sequence of words, so the output can contain inaccuracies. Google calls it critical to manually fact check and review all AI-generated content for accuracy and trustworthiness before publishing, and says that review also covers titles, meta descriptions, structured data and image alt text.

The same guidance warns that using AI to generate many pages without adding value for users may violate Google’s spam policy on scaled content abuse. That policy, last updated August 28, 2026, defines the problem as many pages generated for the primary purpose of manipulating search rankings and not helping users, no matter how they are created. For practices, the familiar version is a set of near-identical city pages or condition pages that differ only by a place name or a diagnosis. Our multi-location practice SEO guide covers how to build location pages that are genuinely different.

Two risks are specific to health content:

Invented facts and sources. A draft can state a statistic, cite a study or name a guideline that does not exist, or that says something different. Every number, every “studies show” and every citation has to be checked against the source itself.
Drift into medical advice. A general explainer can quietly turn into instructions for a specific reader: what to take, how much, when to stop. A clinician should review anything that reads that way. Our guide to writing a medical practice blog covers where general education ends and advice begins.

Google also suggests telling readers how content was created when that would help them, for example by explaining how automation was used. Many practices handle this with a short line on each post naming the clinician who reviewed it, which is also the most useful signal for a patient.

AI-generated testimonials and reviews

This is where the rules are most direct. The FTC’s rule on consumer reviews and testimonials, 16 CFR Part 465, took effect on October 21, 2024. Section 465.2 makes it an unfair or deceptive practice for a business to write or create a consumer review or testimonial that misrepresents that the reviewer exists, or that the reviewer actually had experience with the business. The FTC’s announcement of the final rule named AI-generated fake reviews as an example of a review by someone who does not exist.

In practice, that rules out:

Generating “patient” reviews or quotes and posting them on your website, social pages or a review platform.
Asking a tool to “polish” a real patient’s short comment into a longer testimonial that adds experiences, results or feelings the patient did not express. Section 465.2 also covers a testimonial that misrepresents the experience of the person giving it.
Using a generated face, name or voice to present a testimonial as if it came from a real patient.

The FTC Endorsement Guides point the same way. Section 255.1 says endorsements must reflect the honest opinions, findings, beliefs or experience of the endorser, and that an endorser who claims to use a product must have been a bona fide user of it. An example in section 255.2 says that when an ad gives the impression of showing real customers and actors were used instead, that fact should be clearly and conspicuously disclosed; a generated person is no more a real patient than an actor is. Section 255.2(b) adds that a testimonial about a key result will be read as typical, so the practice needs evidence that it is, or a clear statement of what patients can generally expect.

A note on the December 2025 order. In 2024 the FTC settled with a company whose AI writing service could produce customer reviews and testimonials. On December 22, 2025, the FTC reopened and set aside that order, saying the complaint’s facts did not support a violation by the tool’s seller and that the order unduly burdened AI innovation. That decision was about the seller of a general tool. It did not change Part 465, which still prohibits a business from creating or disseminating fake reviews and testimonials about itself. Our guide to patient testimonials and before-and-after photos covers how to collect real ones properly.

AI-generated and AI-edited images

Images are where AI is most tempting for practices: no photo shoot, no model releases, any setting you like. The rules depend on what the image is presented as.

Illustrative images. A generated picture of a calm waiting room or a person stretching, used as decoration on a general page, is usually low risk, as long as it does not imply something untrue. It should not suggest a result, a facility you do not have, or a real patient.
Before-and-after images. A generated or altered before-and-after image presented as a real patient’s result is a false depiction of results. It runs into the same FTC principles as a fabricated testimonial, and the Google Ads Manipulated media policy, which does not allow manipulating media to deceive, defraud or mislead. If you use a simulation at all, label it clearly as one and never present it as an outcome.
Real patient photos edited with AI. Retouching, relighting or enhancing a real patient’s photo can change what the photo shows about the result. The FTC’s typicality and net impression principles apply to the edited version, not the original.
Images of your clinicians and team. A generated headshot of a provider who does not work at the practice, or an invented team, misrepresents who will treat the patient.

Illustration: a med spa wants a before-and-after carousel for a new skin treatment but has only two consented patient photo sets. Generating more examples would misrepresent results. Showing the two real sets, with a statement of typical results and a note that individual results vary, stays on firmer ground.

Patient photos raise HIPAA questions too. Full face photographs and any comparable images are on the HIPAA list of identifiers in 45 CFR 164.514(b)(2)(i)(Q), and using a patient’s protected health information for marketing generally requires a written authorization under 164.508(a)(3). Running a patient’s photo through an AI editor does not remove that need, and uploading the photo to an outside tool is a disclosure to that tool’s vendor.

Patient information and AI tools

The fastest way to create a HIPAA problem with AI is not the output; it is the input. Pasting a patient’s message into a writing tool to draft a reply, uploading a chart note to summarize it for a case study, or feeding appointment data into a tool to write follow-up emails all send protected health information to a vendor.

Under 45 CFR 164.502(e), a covered entity may let a business associate create, receive, maintain or transmit protected health information on its behalf only with satisfactory assurance, in a business associate agreement, that the vendor will safeguard it. A consumer writing tool used under standard terms is usually not set up for that. The simplest rule for marketing staff is to keep patient information out of general AI tools entirely, and to write marketing from scratch or from de-identified, invented scenarios that are labeled as illustrations. Our guide to AI chatbots on a practice website covers what to ask a vendor when a tool does need to handle patient messages.

Synthetic voices and video

Generated video and cloned voices are now easy to make. The same principles apply. A synthetic spokesperson presented as a real patient is a fake testimonial. A synthetic version of your own physician needs that physician’s agreement, and should say only what the physician would say and can support. A voice or face that resembles a well-known person can suggest an endorsement that does not exist, and the Google Ads Misrepresentation policy does not allow ads that make it seem a business is supported by another brand, organization or government entity when it is not. Outbound calls with AI voices, synthetic clinicians and state voice laws are covered in our guide to AI voices and avatars in practice ads.

A review process for AI-assisted marketing

Illustration: a three-provider chiropractic practice uses AI to draft its monthly blog post, social captions and ad copy. A process like this keeps the speed while catching the problems.

No patient information in. Drafts start from the topic, public sources and your own notes. Never paste in patient messages, records or photos unless the tool is covered by a business associate agreement and that use is approved.
Check every fact against its source. Numbers, studies, guidelines and quotes are verified at the source. If it cannot be found, it comes out.
Strip unsupported health claims. Words like proven, cure, permanent, guaranteed or painless need evidence the practice actually has. Most should simply be removed.
Clinician review. A licensed clinician reads anything that discusses symptoms, conditions or treatment, and their name goes on the page as reviewer.
No generated patients. No generated reviews, testimonials, patient quotes, patient faces or before-and-after results. Real ones come only with consent and a HIPAA authorization.
Label simulations. Any illustrative or simulated image that could be mistaken for a real result is labeled clearly.
Check the platform rules. Ads are checked against Google Ads policies, including Misrepresentation, before they run; our Google Ads for practices page explains how we manage that review, and our guide to writing healthcare practice ad copy without breaking advertising rules covers the policies that most often stop health ads.
Keep a record. Save the final version, who reviewed it and the sources used. It makes later questions much easier to answer.

How we handle this

We use AI tools for research and first drafts in some of our content work, never as the final word. Every factual claim is checked against a primary source, health statements are kept general, nothing we publish presents a generated person as a patient, and patient information does not go into general AI tools. Our work is built with HIPAA in mind, and we confirm in writing what a BAA covers before any patient information flows through a tool; if it is not covered, we do not route patient information through it. See our healthcare SEO service for how we approach content for practices.

Is it legal for a medical practice to use AI to write marketing content?

Using AI as a drafting tool is not prohibited by the FTC rules or Google’s guidelines discussed here. What matters is the result: the content must be accurate, health claims must be supported by evidence, and testimonials and reviews must be real. The practice is responsible for what it publishes.

Does Google penalize AI-generated content?

Google says appropriate use of AI or automation is not against its guidelines. Generating many pages mainly to manipulate rankings, with little value for users, can violate its scaled content abuse spam policy no matter how the pages are made. Google also says AI content should be fact checked before publishing.

Can we use AI to write patient reviews or testimonials?

No. The FTC rule at 16 CFR 465.2 prohibits a business from creating reviews or testimonials that misrepresent that the reviewer exists or had real experience with the business, and the FTC named AI-generated fake reviews as an example.

Can we use AI images on our website?

Generic illustrative images are usually fine if they do not imply something untrue. Do not present generated or altered images as real patient results, real patients or real team members, and label any simulation clearly.

Do we have to disclose that content was made with AI?

The rules discussed here focus on whether content is deceptive, not on how it was made. Google suggests explaining how automated content was created where that helps readers. A disclosure is required where the content would otherwise mislead, such as a simulated image that could be taken for a real result.

Can staff paste patient emails into an AI tool to draft replies?

Not into a general tool without a business associate agreement that covers that use. Sending protected health information to a vendor that handles it on your behalf requires the assurances in 45 CFR 164.502(e). Ask counsel about your specific tools.

Which state laws require telling patients they are talking with AI, and wording for chat, voice and letters, are in our guide to state AI disclosure laws for patient messages.

AI-generated content and images in healthcare marketing: the rules that already apply