Is an AI receptionist HIPAA compliant? A med spa checklist

At 11 pm, a patient sends your clinic’s Instagram a photo of her swollen lip.

Your AI front desk is the first to see it. Whether that moment is a privacy problem has very little to do with the AI. It depends on the inbox she chose, and on paperwork you signed months ago.

So here is the short answer. No AI receptionist is HIPAA compliant by itself. HHS says it doesn’t certify technology products.

What you can check is whether every vendor that handles patient health information signs a business associate agreement (a BAA). Then check what happens on the channels that won’t.

Instagram is one of those channels. I’ll come back to the swollen lip at the end.

First, check whether HIPAA covers your clinic

HIPAA applies to covered entities and the vendors who work for them. A health care provider is a covered entity if it sends health information electronically for standard transactions, such as billing insurance.

Plenty of med spas are cash only, so the answer can be less obvious than owners expect. Ask your lawyer. State privacy laws can apply either way.

My advice is simpler. Treat patient health details as if HIPAA applies. Your patients assume it does.

What the rules actually say

Here is what the official sources say today, in plain words. Each one is linked at the end.

A vendor that creates, receives, keeps or sends health information on your behalf is a business associate, according to HHS. You need a signed BAA with it before it touches that information.

HHS now spells out the AI case. Its list of business associate examples includes a vendor chatbot that handles “symptom assessment, medical reminders, and appointment scheduling”. That is close to a description of an AI front desk.

Storing it counts. A cloud company that keeps your patients’ information is a business associate, HHS says. That holds even if the data is encrypted and the company can’t read it.

Messaging patients is allowed. HHS says providers can email patients about their care, with reasonable safeguards such as limiting what is said. Its FAQ is about email. I would apply the same care to texts, and check that with your lawyer.

Meta publishes no BAA for Instagram or Facebook messages, and we could not find a way to get one (checked 27 October 2026). So clinical details should stay off DMs, and your AI front desk should know how to move a patient somewhere safer.

Questions that protect patient health data

An AI front desk is never one vendor. It is an AI model, a texting service, a database, hosting and your booking software, working together. Every one of them can touch patient information.

Say you own a med spa. Call her Elena. She is made up for this article. Her vendor tells her the system is “HIPAA compliant” and stops there.

That sentence answers nothing.

• Which vendors in this system will touch patient health information? Name every one.
• Has each of them signed a BAA, or will it before launch? Who is the BAA with, me or you?
• Are the accounts that hold patient data in my clinic’s name, or in yours?
• Who can read the conversations, and is every access logged?
• Can I export every conversation, and delete it when I need to?

A good answer is a list of named vendors and a yes next to each one. A red flag is “we are HIPAA certified”. HHS’s Office for Civil Rights says it does not endorse or certify technology products.

Questions that protect your DMs

Your Instagram inbox is where the most private messages often arrive. It is also the inbox with no BAA behind it.

Elena can’t stop patients from DMing. She can decide what the AI does when one starts sharing symptoms there.

• What does the AI say when someone shares clinical details in a DM?
• Can it move that patient to a channel covered by a BAA, such as a secure link or a call from the nurse?
• Does it ever give medical or dosing advice? It shouldn’t.
• What happens with urgent symptoms, at 11 pm on a Sunday?

A good answer to the last question is specific. Urgent symptoms page your on-call provider, and the patient is told to call 911 in an emergency. Everything else clinical goes to your nurse.

Questions that protect your win-back list

Bringing past patients back by text is where the money is. It is also where a clinic can get into trouble fastest.

Two sets of rules meet here. HIPAA has its own definition of marketing, and messages about your clinic’s own services are treated differently from paid ones. Ask your lawyer where your win-back texts fall.

The texting rules are stricter. Under the Telephone Consumer Protection Act, automated marketing texts need prior express written consent. Since 11 April 2025, a patient can revoke it in any reasonable way, and you have ten business days to honor it.

One more change is coming. An FCC rule taking effect on 25 November 2026 lets a sender name one opt-out method, such as replying STOP. Ask your vendor whether it is ready.

• Do win-back texts go only to patients who opted in? Where is that consent recorded?
• Does every text honor STOP, and does a plain “please stop texting me” count too?
• Does anything health related, like a treatment name, appear in a marketing text?

A good answer: opt-in only, consent stored with the patient record, and STOP honored everywhere. A red flag is “we text your whole patient list”.

Questions that protect patient trust

Patients forgive an assistant. They don’t forgive finding out later that “Jess at the front desk” was software.

Some states now write this down. Since 2019, California has required a bot that tries to sell something to say it is a bot. Since 2025, its AB 3030 has covered clinics that use generative AI for messages about clinical information. Those messages need an AI disclaimer and a way to reach a person.

Booking messages are outside AB 3030, and a message a licensed provider reviews is exempt. Utah requires regulated professions to disclose generative AI up front when it collects health data. Texas has required providers to disclose AI used in health care services since 1 January 2026.

• Does the AI say it is the clinic’s assistant, up front, in its first message?
• Can a patient reach a person whenever she asks?
• Which state disclosure rules have you checked for my clinic?

Back to the swollen lip

Here is how that 11 pm DM should go. The AI doesn’t guess at what is wrong. It says it is the clinic’s assistant and that it is getting a provider now. It tells her to call 911 if this is an emergency.

Then it pages the on-call provider and asks the patient to keep medical details off Instagram. The provider takes it from there, on a channel the clinic controls.

Nothing about that is clever. It is a set of decisions made before launch, written down and tested.

HIPAA compliance is not a feature you buy. It is a list of decisions you can check.

How we handle it

Here are our answers to the list. Patient data stays in accounts you own, not ours. Where patient health data is involved, we use vendors that offer a BAA, and we tell you which channels can’t before we build.

It never gives medical or dosing advice. It tells patients it is the clinic’s assistant, up front. Win-back texts go only to patients who opted in, with STOP to opt out. Every conversation is logged, searchable and yours to export.

We are new to med spas, and the clinic scenes on our page are labelled examples. What the front desk covers is on our AI front desk for med spas page. What the after-hours gap costs is in med spa missed calls and DMs.

If you want the checklist answered for your own clinic, book the free call. The audit after it is free too.

Sources

• HHS, Covered Entities and Business Associates (last reviewed 21 August 2024): https://www.hhs.gov/hipaa/for-professionals/covered-entities/index.html
• HHS, Business Associates FAQ (last reviewed 21 October 2026): https://www.hhs.gov/hipaa/for-professionals/privacy/guidance/business-associates/index.html
• HHS, Sample Business Associate Agreement Provisions (2013): https://www.hhs.gov/hipaa/for-professionals/covered-entities/sample-business-associate-agreement-provisions/index.html
• HHS, Guidance on HIPAA and Cloud Computing (last reviewed 23 December 2022): https://www.hhs.gov/hipaa/for-professionals/special-topics/health-information-technology/cloud-computing/index.html
• HHS, FAQ 570, email with patients (last reviewed 26 July 2013): https://www.hhs.gov/hipaa/for-professionals/faq/570/does-hipaa-permit-health-care-providers-to-use-email-to-discuss-health-issues-with-patients/index.html
• HHS, Marketing guidance: https://www.hhs.gov/hipaa/for-professionals/privacy/guidance/marketing/index.html
• 47 CFR 64.1200, TCPA consent rules: https://www.ecfr.gov/current/title-47/section-64.1200
• FCC 24-24, revoking consent, in effect 11 April 2025: https://docs.fcc.gov/public/attachments/FCC-24-24A1.pdf
• Federal Register, FCC 26-67, opt-out methods, effective 25 November 2026: https://www.federalregister.gov/documents/2026/10/26/2026-21821/enhancing-the-ability-of-consumers-to-control-the-robocalls
• California Business and Professions Code 17941 (SB 1001): https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=BPC&sectionNum=17941
• California AB 3030 (2024), Health and Safety Code 1339.75: https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202320240AB3030
• Utah SB 226 (2025): https://le.utah.gov/Session/2025/bills/enrolled/SB0226.pdf
• Texas HB 149 (2025): https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.htm

Written by

Shahzaib Shoaib — Founder and lead engineer

Profile →