Medical, dental and mental health · Compliance

Can doctors, dentists and therapists use AI to write blog posts?

Yes. Nothing in either rulebook asks who typed it.

A page on your practice’s website is judged on what it claims and on who can be recognised in it. Neither test changes because software wrote the first draft.

Two rulebooks reach one. HIPAA governs what you may say about the people you treat. The FTC governs what you may promise about a treatment. Neither asks who held the pen, and neither moves an inch of the answering off the practice.

← Articles yourpractice.com
Ready

Draft · What a first appointment actually involves

Most people arrive braced to explain everything at once. The first visit is mostly the other way round: working out what you want to be different, and whether this is the right room for that.

Bring nothing. If you keep a list of your medications, have it on your phone.

Report Card 46 checks run
89
Overall

Strong · one thing to look at

Overall 89 · publishable bar 80

Compliance · gate Pass · 7/7
A sample draft, checked and sitting still. Nothing on this screen has published, and the next click belongs to a clinician at your practice.
What follows from it

Three things do not move, whoever does the typing.

01

A clinician answers for it

A post about a condition is a clinical statement about it. No vendor and no tool takes that on for you.

02

Patients stay out of it

Unless one signed an authorization for this exact use. There is no version of this rule that ends in “probably fine”.

03

Someone reads it first

Before it is live, and not whoever holds the website login. So the useful thing is to make it quick and hard to skip.

01 · Responsibility

Name the person, and make them clinical.

This reads as obvious until you try to name them. A post that says what a procedure feels like, how long recovery takes, or when a symptom is worth worrying about is a clinical judgement, so whoever approves it has to be able to defend it out loud. In a group that is the clinician who owns the service line.

The usual failure is not a practice letting software publish unread. It is the sign-off drifting to whoever runs the website, because they hold the login. Right person to press the button, wrong person to decide the recovery timeline is right.

What a sign-off has to carry
Approved byA named clinician, not “the practice”
DateThe day it went live
VersionThe one that published, not Tuesday’s
Patient contentNone, or the signed authorization beside it
The last row is the one nobody writes down, and the only one anybody asks for.
02 · The patient in the draft

You can tell that story out loud. Publishing it is a different act.

The case you would describe at a study club, name left out, is the same case a blog post turns into a disclosure. A published page is permanent, searchable and open to everyone, and the details that make a story worth reading are the ones that identify the person. A town, an age, an unusual presentation, the month. Three of those together is a patient.

18Identifiers the privacy rule lists. A name is one.
6 yearsHow long a signed authorization has to exist.
BeforehandSigned before you publish, never chased afterwards.

What has to be true first

  • The patient signed an authorization.
  • Written for this use, not the intake pack.
  • Filed where you can find it again.
  • They can revoke it, and you can take the page down.

Four lines, and the reason most practice blogs should hold no patients at all.

What practices reach for instead

  • “I changed the details.” Stripping identity is a defined standard, not an edit. A changed name beside the same town, age and procedure changes nothing.
  • “They were happy for me to.” Said in the chair is not signed for this. The rule wants a signature covering the use you are making.
  • A review they left you. They chose to publish that. Repeating it beside what you treated them for is you publishing it.
  • A photo with the face cropped. The crop takes the face. It leaves the tattoo, the ward, the window, the date stamp.

None of this is new, and none of it is caused by software. What software changes is how easily a patient reaches a draft at all, because the fastest way to make a post feel specific is to reach for a real case.

03 · Review

The third one is the only one software takes work off.

The first two are yours and stay yours. The third is a reading job, which is where Verand comes in and the only place it does. It drafts the post from what your practice has told it, checks it against the rules you are subject to, marks what would be a problem and hands it to a clinician. It cannot publish, and no setting changes that.

Draft · What a first appointment actually involvesWaiting for a clinician
Checked againstThe rules your practice is subject toDone
Flagged for youOne outcome claim, one uncited figureHeld
PublishedNo. Not until a clinician here approves it
Verand stops here. A clinician presses publish.
Step 01

Read it as a clinician

Is it right, current, and something you would say in the room? Nothing outside your practice can answer that.

Step 02

Hunt for the patient

Any case, any photograph, any “a patient of ours”. If one is there on purpose, the authorization is filed first.

Step 03

Every figure to a source

And a date beside it. A superseded recommendation reads exactly like a current one.

Step 04

Read the promises sideways

Cut anything guaranteeing an outcome, calling a result permanent, or calling a procedure painless.

Step 05

Disclaimer on the live page

Confirm it appears as the page will publish, not only in the draft.

Step 06

Approve, file, publish

In that order, by name and date. Afterwards it stops happening.

What to expect

Four ways a health draft goes wrong, and where to look.

Pattern 01

A patient who was never there

Ask for a specific example and you get one: an age, a presentation, an outcome. Nobody real is in it, so it is not a privacy problem. It is an invented clinical claim under a clinician’s name, which is the other one.

Pattern 02

Outcome language nobody chose

Models have read a great deal of clinic marketing, most of it by people your board does not reach. So absolutes arrive in a draft about a routine procedure, in a sentence nobody wrote on purpose.

Recovery is quick, and most patients get permanent results from a single visit.

Flagged for you, and on the hard tier there is no override.

Pattern 03

Claims about your own practice

How long you have been in practice, which board certified you, how many of a procedure you have done. Software cannot verify a word of it and writes it anyway, because that is the sort of sentence that belongs there.

Pattern 04

Guidance that has since moved

Screening intervals, age bands, dosing, the season’s vaccine advice. A draft states the version it learned with the confidence of the version in force, and on a health page that error has the longest tail.

The first three are why a clinician reads the draft. The fourth is why every figure carries a source and a date, which software is better at than people.

One limit, said plainly.

The automatic check for patient detail is narrower than you would want. It reads the finished draft for the mechanical giveaways, a Social Security number or a written-out date of birth. It cannot read a paragraph and tell you the person in it is recognisable; that needs someone who knows the patient and the town. Step two of the review above exists for exactly that reason, and we would rather write it down than let you assume otherwise. Verand does not check your state board’s own rules yet either, so what your board bans and the pack does not, you add yourself.

Questions practices ask

The four that come up every time.

Do I have to tell patients a post was drafted with AI?

Nothing in the privacy rule or the FTC’s health-claims guidance asks for a byline naming the software. Both ask whether the page is accurate, supportable and free of anyone’s health information. Plenty of practices disclose anyway; treat that as a decision for you and your counsel, not a requirement you are failing.

Can I publish a patient story if the patient is pleased about it?

Only with a signed authorization written for that use, kept six years, and only while they have not revoked it. Being pleased is not the same as having signed, and neither is a review they left elsewhere. Verand will not catch a story told in prose, so it goes in by hand after the gate.

Does a clinician have to review it, or can my practice manager?

Whoever can judge the clinical content, which means a clinician. Your practice manager can own the queue and press publish, but should not be the one deciding the recovery timeline is right. Verand hands it over with the questionable sentences marked and the sources attached, which is what turns two hours into five minutes.

Could Verand publish while I am with patients?

No, and no plan, setting or connector changes it. Every draft stops at a person at your practice. The connector cannot publish, cannot clear a compliance block and cannot delete content. If nobody reads it, it does not go live.

Read next

Two questions this page leaves to their own pages.

Sources: the HIPAA Privacy Rule’s authorization, identifier and documentation requirements, and the FTC’s health-claims guidance. Validated against HHS and FTC rules. AI-researched and operator-reviewed. Your counsel confirms applicability. Not legal advice. 45 CFR §164.508 · 45 CFR §164.514(b)(2) · 45 CFR §164.530(j)(2) · FTC Act §5 · FTC Health Products Compliance Guidance

The answer is yes. The reading stays clinical.

Seven days, every feature unlocked, one click to cancel. Your first article is drafted during setup, and it stops where this page says.

Start for Free See pricing →

Every draft stops at a person at your practice. There is no setting that changes it.

Verand

Content built to rank in Google and get cited by ChatGPTPerplexityGeminiClaude, with every claim checked before it goes live.

support@verand.ai

© 2026 Verand. All rights reserved. TermsPrivacyAI policyAccessibilitySecurity
Not legal advice. Compliance packs are AI researched and operator reviewed.