Technology AI

Using AI to draft review replies carefully

How plumbing companies can use AI to draft review replies without breaking FTC review rules or exposing customer details, with prompts and a workflow.

AssistCall research teamPublished Reviewed by Faruque Amin, VP of Sales and Marketing7 min read

On this page
  1. Why replies matter more than they seem
  2. What AI does well here, and what it does not
  3. The honesty rules
  4. The privacy rules
  5. A reply workflow
  6. Prompts by review type
  7. Asking for reviews without crossing the line
  8. Should you say a reply was drafted with AI?
  9. Example: a late-arrival review
  10. Frequently asked questions
  11. References
Technology AI: Using AI to draft review replies carefullyArticle coverTECHNOLOGY AIUsing AI to draftreview repliescarefullyAva for PlumbersInsightsTechnology AI: Using AI to draft review replies carefullyArticle coverTECHNOLOGY AIUsing AI to draftreview repliescarefullyAva for Plumbers

Key takeaways

  • 82% of US adults at least sometimes read reviews before a first purchase, and 54% of review readers pay more attention to extremely negative ones[1].
  • AI is useful for a calm first draft; a person must check every fact, remove private details and approve every reply.
  • The FTC's 2024 rule bans fake reviews, including AI-generated ones, and incentives tied to positive sentiment[2].
  • Never confirm a customer's address, job details or bill in a public reply.

Why replies matter more than they seem

Most homeowners look you up before they call, and many read the replies as closely as the reviews. Pew found that 82% of US adults at least sometimes read online ratings or reviews before buying something for the first time, and 40% nearly always rely on them[1]. Readers weigh the bad more than the good: among people who read reviews, 54% pay more attention to extremely negative reviews, compared with 43% for extremely positive ones[1].

How US adults use online reviewsPew Research Center, 2016.
How US adults use online reviewsBar chart. At least sometimes read reviews before a first purchase: 82%; Nearly always rely on reviews when buying: 40%; Review readers who pay more attention to extremely negative reviews: 54%; Review readers who pay more attention to extremely positive reviews: 43%.At least sometimes read reviews before a first purchaseAt least sometimes read reviews before a first purchase: 82%82%Nearly always rely on reviews when buyingNearly always rely on reviews when buying: 40%40%Review readers who pay more attention to extremely negative reviewsReview readers who pay more attention to extremely negative reviews: 54%54%Review readers who pay more attention to extremely positive reviewsReview readers who pay more attention to extremely positive reviews: 43%43%How US adults use online reviewsBar chart. At least sometimes read reviews before a first purchase: 82%; Nearly always rely on reviews when buying: 40%; Review readers who pay more attention to extremely negative reviews: 54%; Review readers who pay more attention to extremely positive reviews: 43%.At least sometimes read reviews beforea first purchaseAt least sometimes read reviews before a first purchase: 82%82%Nearly always rely on reviews whenbuyingNearly always rely on reviews when buying: 40%40%Review readers who pay more attentionto extremely negative reviewsReview readers who pay more attention to extremely negative reviews: 54%54%Review readers who pay more attentionto extremely positive reviewsReview readers who pay more attention to extremely positive reviews: 43%43%

Source: Pew Research Center, 2016[1]

Show the data
ItemValue
At least sometimes read reviews before a first purchase82%
Nearly always rely on reviews when buying40%
Review readers who pay more attention to extremely negative reviews54%
Review readers who pay more attention to extremely positive reviews43%

That makes your reply to a one-star review one of the most-read things your company writes. A defensive or careless reply confirms the reviewer's story. A calm, specific reply shows the next customer how you behave when something goes wrong. Only around one in ten Americans nearly always post reviews themselves[1], so each review, and each reply, carries extra weight.

Replying is also time-consuming, especially for owners who answer reviews at night after a long day. That is where AI drafting helps, as long as it is used for wording and not for judgment.

What AI does well here, and what it does not

AI is good at:

  • Taking the heat out of an angry first draft written by the owner.
  • Keeping replies short, polite and consistent in tone.
  • Varying wording, so twenty thank-you replies do not read identically.
  • Suggesting a structure for a difficult reply: acknowledge, explain briefly, offer a next step.

AI is bad at, and should never be trusted with:

  • Knowing what actually happened on the job; it will fill gaps with plausible fiction.
  • Judging whether a reply admits fault in a way your insurer would object to.
  • Recognizing private details that should not appear in public.
  • Deciding whether a review is fake or from a competitor.

The honesty rules

The Federal Trade Commission's final rule on consumer reviews and testimonials, announced in August 2024, bans creating or selling fake reviews, including AI-generated fake reviews[2]. It also bans offering compensation or incentives conditioned on a review expressing a particular sentiment[2], and bans using unfounded legal threats, physical threats or intimidation to suppress negative reviews, with civil penalties possible for knowing violators[2].

For a plumbing company using AI near reviews, that translates into clear lines:

  • Use AI to draft your replies, never to write reviews for your company, on any platform, by anyone.
  • Do not ask staff, family or friends to post reviews, AI-written or not.
  • You may ask every customer for an honest review; you may not offer a discount, gift or entry into a draw only for positive ones.
  • Do not threaten reviewers with legal action in replies or messages; if a review is defamatory, talk to your attorney privately.
  • Do not let AI invent facts in a reply, such as claiming the customer refused a repair if your records do not show it.

The privacy rules

A review reply is public and permanent. It should never confirm personal information, even if the reviewer posted it first. Larger businesses face formal privacy duties; the California Consumer Privacy Act applies to for-profit businesses in California above set thresholds, such as gross annual revenue over $25 million[3]. Every business, of any size, should follow these habits:

  • No addresses, street names or neighborhood details that identify the home.
  • No invoice amounts, payment history or account status.
  • No details about who was home, medical needs, pets or access codes.
  • No technician's full name if it would expose them to harassment; first name or role is enough.
  • Move anything specific to a private channel: please call our office manager at the main number.

Apply the same rule to what you paste into the AI tool. Paste the review text, which is already public, and your short notes on what happened, without the customer's name, address or invoice details.

A reply workflow

  1. Check the record

    Find the job. Read the notes, call summary and invoice so you know what actually happened before you write anything.

  2. Write three facts

    Note in plain words what happened, what you will offer, and how the reviewer can reach you privately.

  3. Draft with AI

    Paste the review and your three facts with a standard prompt. Ask for a reply under 80 words, with no new facts.

  4. Edit for truth and privacy

    Delete anything not in your notes and any personal detail. Check the offer is one you will honor.

  5. Approve and post

    The owner or a named manager approves negative and mixed replies. Positive replies can be approved by the office manager.

  6. Follow up privately

    For negative reviews, make the private contact you promised within one business day.

Prompts by review type

How to handle each type of review, and who signs off.
Review typeReply approachWho approves
Five stars, specific praiseThank them, mention the work type, invite them backOffice manager
Five stars, no textShort thanks; vary wordingOffice manager
MixedThank them, acknowledge the issue, offer private follow-upOwner or manager
Negative, real customerAcknowledge, brief factual context, private next stepOwner
Negative, no matching recordSay you cannot find the job and ask them to contact youOwner
Abusive or off-topicUsually report to the platform rather than replyOwner
  • Prompt for positive reviews: Write a warm reply under 50 words thanking this customer for their review of a drain cleaning. Do not mention any personal details. Sign as the Riverside Plumbing team.
  • Prompt for negative reviews: Draft a calm, non-defensive reply under 80 words. Facts: the technician arrived two hours late because an emergency ran over; we did not call ahead; we are refunding the service-call fee. Do not add any other facts. Invite them to call the office manager. Do not mention their name or address.
  • Prompt for no matching record: Draft a polite reply under 50 words saying we cannot find a job matching this review and asking the reviewer to contact our office so we can look into it.

Asking for reviews without crossing the line

Replies are only half of review management; the other half is getting more honest reviews in the first place, so one bad night does not define your page. AI can help draft the request message, but the rules on incentives and fake reviews apply to the request just as much as to the reply.

  • Ask every customer the same way, not only the ones you expect to be happy. Filtering requests so only satisfied customers are asked can look like review gating.
  • Keep the request short: thank them for the job, give one link, and say honest feedback helps you improve.
  • Send it soon after the job, while the visit is fresh, and only once or twice.
  • Never offer anything that depends on the review being positive, and be careful with any incentive at all; check the platform's own rules.
  • Do not draft a review for the customer to paste, even if they ask you to; offer to tell them what work was done so they can describe it themselves.

Track where reviews come from and how many arrive each month. A steady flow of genuine reviews is the best protection against the outsized effect of a single negative one.

Should you say a reply was drafted with AI?

There is no general rule requiring businesses to label AI-assisted review replies. But people care about knowing: in a 2025 Pew survey, 76% of US adults said it was extremely or very important to be able to tell whether content was made by AI or by people[4]. The practical answer is to make sure every reply is genuinely yours: edited, approved and signed by a real person or by your team. A reply that reads as generic, or that obviously misunderstands the review, will be taken for automation whether it is or not.

Example: a late-arrival review

Example review: Two hours late, nobody called, and I had to take time off work. Example reply after editing: Thank you for telling us, and we're sorry. Our technician was held up on an emergency job and we should have called you to say so; that's on us. We've refunded the service-call fee, and our office manager would like to talk about how we keep you updated next time. Please call the main number and ask for Dana. The Riverside Plumbing team. The owner removed a line the AI had added promising that it would never happen again, because that was not a promise the shop could guarantee.

Frequently asked questions

Is it legal to use AI to reply to reviews?

Yes. Using AI to draft your own replies is not the problem. Using AI to create fake reviews, or offering incentives only for positive reviews, is banned under the FTC's 2024 rule.

Should every review get a reply?

Reply to every negative and mixed review, and to as many positive ones as you can without sounding repetitive. Short, varied replies are better than identical long ones.

Can we reply with details to correct a false review?

Keep the public reply brief and factual without personal details, and move the conversation to a private channel. For a defamatory review, speak to your attorney rather than arguing in public.

Can an AI tool reply to reviews automatically?

Some platforms offer it, but automatic posting removes the human check that catches invented facts and privacy slips. Use drafting, not auto-posting, at least for negative and mixed reviews.

References

  1. Pew Research Center (2016). Online Reviews (Online Shopping and E-Commerce) (opens in a new tab) โ†‘

  2. Federal Trade Commission (2024). Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials (opens in a new tab) โ†‘

  3. State of California Department of Justice, Office of the Attorney General (2026). California Consumer Privacy Act (CCPA) (opens in a new tab) โ†‘

  4. Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) โ†‘

Book a demo

See Ava handle your calls.

Book a short demo. Weโ€™ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo