How a medical or wellness practice should answer a bad online review
A one-star review lands late on a Friday. What you can say, what you must never say, when to say nothing, and how AI can help you draft the reply safely.
- A public reply may never confirm someone was a patient, their condition, or their visit
- Decide whether a reply helps at all before writing one
- AI is safe here only as a drafting aid, with a human checking every word
- A second pair of eyes reads the draft before it posts
It usually starts the same way. A one-star review appears, the name looks like a patient, and the first instinct is to explain the visit from your side. The instinct is natural and it is the one that creates the real risk. For a medical or wellness practice, even a kind, factual reply can cause harm if it confirms that the person was ever your patient.
The rule set is simple. Never reveal patient details. Decide whether a reply helps at all. Use AI only to draft and review, with a human deciding what actually posts.
What may never appear in a public reply
Health privacy rules sit behind every public reply a practice writes. The details a practice may not confirm include the fact that the person is a patient, any condition they were seen for, the date or nature of a visit, and any test, treatment, or billing detail tied to them. That applies even when the reviewer posts those details first. If a patient names their own diagnosis in a review, a public reply that engages with that diagnosis confirms the relationship and the record.
That constraint pushes good replies toward one narrow shape: a short, calm response that invites the conversation offline. No facts about the visit. No rebuttal of the complaint. No apology that confirms the person was treated. Something like: thank you for the feedback, we take concerns seriously, please contact the office directly so we can help.
Decide whether to reply at all
A reply is not owed. A calm, compliant reply signals to future patients that the practice reads feedback and handles it professionally. A defensive reply reads worse than no reply, and a rushed one can read worse than the complaint itself. If the review is a clear service complaint, the value of a short reply is usually worth it. If it is abusive, contains private health details, or reads like a legal dispute in the making, the better move is often to report the review through the platform's own reporting process and stay silent in public.
Sleep on the draft if the review stung. The same words posted a day later read differently from the ones posted an hour after the review lands, and the delay costs nothing.
How to use AI to draft the reply
AI tools such as ChatGPT or Claude are useful here as drafting aids, nothing more. They help with tone, brevity, and consistency across replies, and they help a practice write the same compliant response every time. They never decide what posts. The owner or manager reads every word before it goes up.
- 01Set the guardrail first
Open your prompt with the rule: the draft may not confirm the reviewer was a patient or reference any condition, visit, or record.
- 02Describe only the situation
Give the tool the complaint's general topic, such as wait times or front desk service, with no names, dates, or health details.
- 03Ask for three short versions
Request a two to three sentence reply in a calm, professional tone, written so it could apply to any reviewer.
- 04Check each against the rule list
Read every draft for anything that implies the person was treated, what they were seen for, or when.
- 05Route the complaint offline
Keep the line that invites the person to contact the office directly.
- 06Get a second pair of eyes
Have someone else at the practice, or a trusted outside reviewer, read the final draft before it posts.
Keep drafts and prompts clean
The prompt itself should carry no patient information. Describe the complaint in general terms only. Free and paid AI tools keep and process what you type, so anything with a name, a date of birth, or a diagnosis does not belong in the box. A prompt that says a reviewer complained about a long wait is safe. A prompt that says why that person was in the office is not.
A short standing policy helps. One sentence stating what a public reply may and may not include, kept where the whole team can see it, keeps the practice from relearning the rules each time a review lands.
Where a concierge fits
Some owners want a second pair of eyes without adding headcount. New York AI Lab offers an AI Concierge: one of its people learns the business and works alongside the owner, with two 1-hour calls a month and voice messages and email between calls. A concierge can review draft replies to reviews against the practice's rule list between calls, and help set up the prompt template so the drafting step is the same every time.
