Using ChatGPT to Respond to Dental Practice Reviews: Templates, Tone, and What Not to Say

Most dental practices have somewhere between 30 and 100 Google reviews. A significant number of those, especially the negative ones, go unanswered. Not because the practice doesn’t care, but because nobody has time. The office manager is scheduling, handling insurance calls, and fielding the front desk. The dentist is in a chair. Writing a thoughtful response to a one-star review at 8pm on a Tuesday is a low-priority task that keeps getting pushed.

That silence costs patients. BrightLocal’s consumer research shows 88% of people read business responses to reviews. An unanswered negative review looks like confirmation of the complaint. An unanswered positive review is a missed connection with someone who just told the internet they liked you.

ChatGPT can draft review responses in seconds. An office manager who currently spends 20-30 minutes a week writing responses (or more realistically, avoiding writing them) can cut that to 5 minutes of reviewing and editing AI drafts. That’s not a marginal improvement. That’s the difference between responding to every review consistently and responding to none of them.

Here’s the risk: ChatGPT does not know HIPAA. Given a one-star review that describes a specific treatment, it will write a response that addresses the treatment. Given a review from “Sarah M.” who mentions her crown fell out, it will write a response that references the crown. Both of these responses could be HIPAA violations. This guide gives you the prompt template that constrains ChatGPT to HIPAA-safe responses, shows you what those responses actually look like across 8 common review scenarios, and tells you what to watch for before you post anything.

Why HIPAA Matters for Review Responses

Most dentists know HIPAA applies to patient records, lab results, and clinical communication. Fewer realize it applies to public review responses, and the failure mode here is subtle enough that it catches smart people off guard.

The rule is simple: a dental practice responding to a public review cannot confirm or deny that the reviewer is a patient. Even if the reviewer names themselves. Even if they describe their treatment in detail. Even if the review is clearly from a real patient whose chart you can pull up right now. The moment your response acknowledges the treatment, the appointment, or any clinical detail, you’ve confirmed they’re a patient and potentially disclosed protected health information in a public forum.

The practical implications:

• Never reference specific treatments mentioned in the review. If someone writes “the filling you put in last week already fell out,” your response cannot say “we’re sorry about your filling” or anything that confirms the procedure happened.
• Never acknowledge specific visits or appointment details. “We’re sorry your last visit didn’t meet your expectations” sounds harmless but confirms the person visited your practice.
• Never address clinical complaints with clinical explanations. Even vague defenses like “our procedures meet the highest clinical standards” in response to a specific complaint imply you know what procedure the reviewer had.
• The safe approach: respond warmly and generically. Thank them (for positive reviews), express concern (for negative ones), and invite them to contact the office directly to discuss their experience. That’s it.

This is exactly where ChatGPT fails without guidance. Its default behavior for a negative review is to be helpful, to address the specific concerns raised. That helpfulness is the problem. The prompt template below constrains it to safe responses.

The Master Prompt Template

This single prompt works for all review response scenarios. Copy it into a shared Google Doc or Notes app where your office manager can access it. Replace the bracketed fields for each review.

You are drafting a response to a Google/Yelp review for [Practice Name], a general dentistry practice.

CRITICAL RULES, follow these exactly:
- NEVER confirm or deny that the reviewer is a patient
- NEVER reference specific treatments, procedures, or clinical details, even if the reviewer mentions them
- NEVER acknowledge specific dates, appointments, or staff members by name
- Keep the response under 100 words
- Tone: professional, warm, genuine, not corporate or defensive
- For negative reviews: express concern, invite them to contact the office directly at [Phone Number]
- For positive reviews: express gratitude, keep it brief and personal
- Sign as [Doctor's Name or "The [Practice Name] Team"]

The review says: "[paste review text here]"

The review rating is: [1-5 stars]

Draft a response.

A few notes on why the prompt is structured this way:

The HIPAA rules come first. ChatGPT tends to weight earlier instructions more heavily than later ones. If the rules are buried at the end of the prompt, they’re more likely to be overridden by the model’s instinct to be helpful. Front-loading them makes compliance more reliable.

The 100-word limit prevents over-sharing. The longer the response, the more likely the AI drifts into specifics. Short responses are also better responses. Nobody wants to read a 300-word reply to a Google review.

“Not corporate or defensive.” Without this instruction, ChatGPT’s default tone for negative reviews sounds like it was written by a PR agency. “We take all patient concerns very seriously and strive to provide the highest quality of care.” Nobody believes that sentence. The instruction pushes it toward something that sounds like a real person wrote it.

The phone number in negative review responses. Including the actual phone number in the prompt is intentional. It means the AI will write it directly into the draft, and you don’t have to add it manually before posting.

8 Example Responses

These prompts were tested in ChatGPT (GPT-4o) and Claude 3.5 Sonnet. We did not deploy this workflow in a live practice or respond to real patient reviews. The reviews below are written examples, not real submissions.

1. Five Stars: General Praise

Scenario: A happy patient, nothing specific, just a positive overall experience.

“Dr. Miller and the whole team are fantastic. I’ve been coming here for three years and always leave feeling like they actually care. Five stars every time.”

Rating: 5 stars

ChatGPT output:

Thank you so much. This genuinely made our day. We’re glad you’ve felt that care over the years, and we look forward to seeing you again soon.

• Dr. Miller and the Team

Assessment: Ready to post. Short, warm, personal without referencing any clinical details. The “made our day” phrasing sounds human. No changes needed.

2. Five Stars: Praise for a Specific Team Member

Scenario: The reviewer names a hygienist and calls out their technique.

“I always ask for Jamie when I come in. Best hygienist I’ve ever had: gentle, thorough, and actually explains what she’s doing. Ten out of ten.”

Rating: 5 stars

ChatGPT output:

Thank you for the kind words. It means a lot to the whole team. We’re glad every visit leaves you feeling well taken care of. See you at your next appointment!

• The Riverside Dental Team

Assessment: Ready to post. The prompt’s instruction to avoid naming staff members led the AI to respond without echoing “Jamie,” which is conservative but safe. Hygienist names are not protected health information, so you could choose to add “Jamie will love hearing this” before posting if you want a more personal touch. Either version is HIPAA-compliant.

3. Four Stars: Good Experience, Mentioned a Long Wait

Scenario: Positive review with one complaint about wait time.

“Really like this practice overall. Dr. Chen is great and the office is clean. Took off one star because I waited 25 minutes past my appointment time last visit. Would still recommend.”

Rating: 4 stars

ChatGPT output:

Thank you for the honest feedback, and for recommending us despite the wait. Your time matters, and we hear you. We’re always working to keep appointments running on schedule. We hope to do better next time.

• Dr. Chen and the Team

Assessment: Ready to post. This is a good response because it acknowledges the concern directly (wait times) without getting defensive or making excuses. The AI correctly didn’t reference the specific visit or confirm the reviewer was a patient; it just addressed the category of complaint. “Your time matters” is a bit close to stock phrasing; you could swap it for “We know showing up on time matters” and it would feel more natural.

4. Three Stars: Liked the Dentist, Disliked the Front Desk

Scenario: Mixed review. Clinical experience was good, administrative experience was not.

“The dentist himself is great, really thorough. But the front desk staff were rude when I called to ask about my bill. Three stars.”

Rating: 3 stars

ChatGPT output:

Thank you for taking the time to share this. We’re glad your time in the chair was positive, but we want every part of your experience with us to feel the same way. Please don’t hesitate to reach out at (303) 555-0182. We’d like to make it right.

• Dr. Torres and the Team

Assessment: Ready to post. The AI correctly avoided naming or referencing the front desk staff specifically, and it didn’t confirm which staff member the reviewer interacted with. “Every part of your experience with us” is a clean way to acknowledge the complaint without specifics. The invitation to call is appropriate here since there’s a billing concern worth resolving privately.

5. Two Stars: Billing Complaint

Scenario: The reviewer says they were charged for something their insurance should have covered.

“Went in for a cleaning and was charged $180 out of pocket. My insurance covers cleanings 100%. When I called to dispute it, they were unhelpful. Very disappointed.”

Rating: 2 stars

ChatGPT output:

We’re sorry to hear this experience didn’t meet your expectations. Billing concerns deserve a proper conversation, and we want to help resolve this. Please call our billing team directly at (303) 555-0182. We’ll look into it.

• The Greenfield Dental Team

Assessment: Ready to post. This is the scenario where the HIPAA guardrail matters most. Without the rules in the prompt, ChatGPT reliably responds by addressing the specific billing situation: acknowledging the charge amount, referencing the insurance claim, or explaining coverage policy. All of those responses would confirm the reviewer is a patient and potentially disclose billing information. The constrained version stays generic: “billing concerns deserve a proper conversation.” That’s the correct move.

6. One Star: Clinical Complaint

Scenario: The reviewer says a filling failed shortly after placement.

“Had a filling done here two weeks ago and it fell out yesterday. Called the office and they said they could see me in three weeks. Three weeks! Going somewhere else.”

Rating: 1 star

ChatGPT output:

We’re sorry to hear about your experience. That’s not the standard we hold ourselves to. We’d welcome the chance to make this right. Please call us at (303) 555-0182 and ask for our scheduling team. We’ll do our best to see you sooner.

• Dr. Patel and the Team

Assessment: Ready to post. Read the response carefully: it does not confirm a filling was placed, does not confirm the reviewer is a patient, and does not address the clinical claim. “That’s not the standard we hold ourselves to” addresses the complaint category (a bad outcome and a poor response from the office) without confirming what happened. The offer to see them sooner is appropriate and addresses the scheduling frustration without admitting fault. This is exactly what a constrained AI response should look like on a 1-star clinical complaint.

7. One Star: Angry, Personal Attack

Scenario: The reviewer is hostile and makes accusations about the dentist’s competence.

“Dr. Williams is incompetent. I had to have my work redone by another dentist. Complete waste of money and a scam. STAY AWAY.”

Rating: 1 star

ChatGPT output:

We’re sorry you had such a negative experience. That’s not something we take lightly. If you’d like to discuss what happened, we’re available at (303) 555-0182. We’d genuinely like to understand your concerns.

• The Williams Family Dental Team

Assessment: Ready to post. This is actually the scenario where the AI’s dispassionate tone is the biggest advantage. Most practice owners who receive a review like this want to respond point by point, which is almost always a mistake. The AI gives you a short, professional response that doesn’t engage with the accusations, doesn’t confirm any clinical details, and leaves a door open for direct contact. Add this to your queue and post it without changes. Do not rewrite it to defend the dentist. If the accusations are false, the professional response wins in the court of public opinion. The combative response does not.

8. Fake or Spam Review

Scenario: The review is obviously fabricated. It describes a procedure the practice doesn’t perform or names a patient who doesn’t match any records.

“I came in for a root canal and Dr. Anderson left a tool inside my tooth. I had to go to the ER. This is medical malpractice.”

Rating: 1 star

ChatGPT output:

We take all concerns seriously and want to understand this better. We don’t have a record that matches this description. If you believe you were seen at our practice, please contact us directly at (303) 555-0182 so we can look into this properly.

• Dr. Anderson and the Team

Assessment: Needs editing before posting. The line “we don’t have a record that matches this description” is close to the line. It’s a denial rather than a confirmation, which is safer, but a HIPAA attorney would likely advise against any statement about your records in a public forum. Consider editing to: “We’d like to look into this directly. Please call us at (303) 555-0182 so we can understand what happened.” More importantly: flag this review with Google for removal simultaneously with posting a response. Google will remove reviews that describe events that couldn’t have occurred (e.g., a procedure you don’t offer) or that appear to be spam. To flag a review in Google Business Profile, click the three-dot menu next to the review and select “Report review.” The review removal process can take several days to weeks; posting a professional response in the meantime is the right move.

Common Mistakes to Avoid

Even with the prompt, there are failure modes to watch for before you post.

The AI echoes the reviewer’s name. If the reviewer signs their name in the review text, ChatGPT sometimes writes “Dear [Name]” or “Thank you, [Name]” at the start of the response. Remove it. Addressing a reviewer by name in your response reinforces the association between that person and your practice.

“We take all feedback seriously.” This phrase appears in roughly 40% of ChatGPT review responses without the prompt restrictions, and occasionally sneaks through even with them. If you see it, delete it and replace it with nothing. The rest of the response is better without it. It signals that a template was used.

Vague clinical defense. Watch for phrases like “our procedures follow the highest clinical standards” or “we always prioritize patient safety.” These sound neutral but, in response to a specific clinical complaint, imply you know what procedure was performed. Remove them.

Every response sounds the same. If you run five positive reviews through ChatGPT with the same prompt and post all five responses in the same week, a reader who looks at your Google page will notice. Vary the responses. Regenerate until you get different phrasing, or edit after the fact. Consistency in posting cadence is good; consistency in exact wording looks automated.

Skipping the review step. Read every response before posting. The AI doesn’t know your practice’s voice, your team, or the specific context that might make a response inappropriate for a given review. Five seconds of review before posting is the right habit.

How to Set This Up in Your Practice

Save the prompt template somewhere accessible. A shared Google Doc works. A note in your practice management software works. The goal is that whoever handles review responses can get to it without asking. Paste it, fill in the blanks, copy the output, edit, post.

Train the office manager on the HIPAA reasoning, not just the rules. The prompt has the rules built in, but the person reviewing the output should understand why, so they recognize a HIPAA problem even if the AI produces one. Five minutes explaining the logic is worth more than five minutes explaining the steps.

Batch responses once a day. Check your Google Business Profile notifications each morning. Draft all pending responses at once, review them, post them. Responding to reviews within 24 hours is a good habit and Google’s algorithm rewards consistent engagement with your profile.

This prompt works in ChatGPT (free or Plus), Claude, and Gemini. The template is model-agnostic. If you already use one of these tools for other practice tasks, like drafting patient follow-up emails or marketing content, use the same one here. There’s no meaningful quality difference between models for this use case.

If your practice uses Weave, Podium, or BirdEye for review management, check whether they have built-in AI response drafting. They may do this natively now. The ChatGPT approach is the right starting point if you’re not already paying for one of those platforms, but it’s worth checking your existing tools before adding a new workflow.

Closing

The real win here is consistency. Most practices respond to reviews in bursts: a few in a good week, none for months after that. The inconsistency is visible to anyone who looks at your Google profile, and it signals that review management is reactive at best. AI makes it practical to respond to every review within 24 hours, regardless of whether anyone has time or emotional bandwidth that day. Google’s ranking algorithm for local search factors in review response rate and recency. Consistent responses compound.

One boundary worth being explicit about: review responses are administrative communication, not clinical communication. This is the right use case for AI in your practice: generic, public-facing text that follows clear rules and doesn’t require clinical judgment. (If you are also hiring for your practice, AI handles job posting drafts equally well.) Clinical communication (treatment recommendations, post-op instructions, responses to specific patient health questions) should be written by a human who knows the patient and the case. The line between those two categories is clearer than it might seem, and this is firmly on the right side of it.