A dermatology and cosmetic clinic ran a social campaign promoting a laser treatment. Within an hour, a commenter wrote that one of the clinic's doctors had treated her face last spring and the results were worth every penny.
The clinic's social coordinator, trying to be warm, replied by name: so glad she loved her results, and see you at your follow-up in June. The reply confirmed the woman was a patient, named her treating doctor, and disclosed a future appointment, all in public. The clinic's compliance officer spotted it that afternoon and had it removed, then spent the next week deciding whether it needed to be reported.
Replies Are Where Social Media Gets Risky
Healthcare marketers usually focus on what they post. The bigger risk sits in how they respond.
A patient can say whatever they like about their own care. The practice can't confirm it. Regulators have fined medical and dental practices for disclosing patient information in replies to online reviews, even when the patient raised the topic first.
The clinic now uses a short set of approved replies. For praise, a general thank-you with no mention of treatment, doctors, or appointments. For complaints, an invitation to call the practice manager directly, with a phone number. Nothing that confirms or denies the person is a patient.
The coordinator who wrote that first reply helped draft the new templates. She'd never been told the rule.
Direct Messages With Photos
The campaign also brought a wave of direct messages. Some asked about pricing. Others included photos of moles, rashes, and scars, with questions like "is this something I should worry about?"
No clinic should assess skin conditions in a social inbox. Messages there aren't protected the way medical records are, and a casual reply can read as a diagnosis. The clinic set an automatic response for messages containing images: thanks for reaching out, we can't review photos or medical questions here, please call or book a consultation. Staff then follow up with the booking link only.
When the Ad Leads to a Phone Call
Many people who see a cosmetic treatment ad want to talk to someone before booking. The clinic's front desk couldn't keep up during the campaign, so it added an AI voice agent to answer overflow and after-hours calls.
Its scope is deliberately narrow. It answers questions about hours, locations, consultation fees, and general price ranges. It explains what a consultation involves. And it books appointments directly into the clinic's HIPAA-compliant scheduling software, which runs under a signed business associate agreement and sends reminders with only the date, time, and location.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
What it won't do matters as much. It never gives medical advice, never predicts results, and never describes a treatment as guaranteed or painless. Any question about a specific medical concern, medication, or existing condition goes to a person or a callback request. Callers asking about an existing appointment must verify their identity before the agent discusses it.
Choosing HIPAA-compliant scheduling software that the voice agent can write to directly mattered more than expected. Earlier attempts that relied on the agent emailing booking requests to staff produced duplicate and lost appointments.
Testing Before the Ad Budget Hits
The clinic tested the agent for two weeks before the campaign launched, using scripted calls written by nurses and the marketing team.
The voice AI evaluation metrics it tracked were specific to a clinic advertising on social. Booking completion rate for callers who wanted a consultation. Correct escalation of medical questions, including tricky ones like "my friend said this laser is safe for pregnant women, right?" Compliance with marketing rules, meaning zero promises about outcomes in any test call. Identity checks before discussing existing appointments. Time to reach a person when the agent transferred. And caller hang-ups.
Each campaign got its own tracking number, so bookings could be tied back to the ad that produced them. After launch, the team kept reviewing its voice AI evaluation metrics weekly and listened to a sample of calls, especially any where the agent transferred late or the caller ended the call early.
Training the People, Too
Tools only help if the people around them understand the rules. Everyone who touches the clinic's social accounts now gets a short privacy training before their first post, with real examples of replies that went wrong. Templates sit beside the inbox, and anything unusual goes to the practice manager before a response goes out.
The Reply That Named a Patient
The patient from the first comment never complained. She came to her June follow-up, and the clinic's manager apologized in person. The campaign went on to fill the laser schedule for three months, and every reply posted after that first one said thank you and nothing more.


