Should You Use AI to Write Customer Reviews?

AI-written and fake are not the same question

Two facts can be true at once. A review can be written without AI and still be fake. A review can also describe a real visit even though the customer used a writing tool.

For a local business, the useful line is not whether a sentence sounds machine-written. It is who had the experience, who supplied the facts and opinion, who controlled the final wording, and whether the platform allows the help they used.

The FTC's Consumer Reviews and Testimonials Rule covers reviews that misrepresent whether the reviewer exists, had the experience, or accurately described that experience. It also prohibits businesses from buying or procuring reviews when the business knew or should have known they were fake. The rule does not make ordinary customers liable simply for writing a review. Its focus is deceptive commercial conduct.

That distinction matters when an owner receives a strangely polished review from a recognizable customer. Odd phrasing is not proof that the visit was invented. Check the facts you can responsibly verify before treating the customer as a fraud.

The rules change with the destination

There is no universal AI-review policy. The same workflow can be acceptable under one set of rules and prohibited under another.

The stricter rule should govern the workflow. If a customer is writing for Yelp, Yelp's ban applies even when the visit really happened. If a testimonial will appear on your own site, do not turn a customer's two-sentence note into a detailed success story they never approved.

• Google: Reviews must reflect a genuine experience and be real and unbiased. Businesses cannot offer incentives, request particular content, pressure customers, selectively ask only for positive reviews, or create reviews from multiple accounts. Serious violations can lead to removed reviews, posting restrictions, unpublished reviews or a warning on the profile.
• Yelp: Yelp says reviewers must write their own review from direct, firsthand experience. It does not allow third-party AI tools to create, draft or revise the review. Yelp also discourages businesses from soliciting reviews in the first place.
• Your own website: A testimonial selected for a business website is advertising, not a neutral review platform. The FTC says businesses publishing testimonials on their own sites do not receive the rule's hosting exemption. The experience, endorsement and any material connection still need to be truthful.

Five review shortcuts, judged plainly

A tool is less important than the work it is being asked to do. These common situations show where the risk changes.

A reliable review program does not need to operate near these boundaries. It needs a consistent invitation and a clean handoff to the customer.

• The business generates ten glowing reviews and posts them from staff accounts. No. The reviewers and experiences are false, and the business is creating the deception.
• An agency writes a review and asks a real customer to paste it into Google. No. The business side is supplying the praise and specific content. A customer's account does not make the review independent.
• An employee reviews the business without disclosing the relationship. No. The relationship is material, and Google treats undisclosed conflicts of interest as rating manipulation.
• A customer writes down what happened, then asks AI to fix spelling. The experience may be real, but the destination still decides whether that assistance is allowed. Yelp says no even for revision. Google's policy does not explicitly settle this narrow case, so the business should not promise that it is approved.
• The owner uses AI to draft a reply to a review. That is an owner response, not a customer review. It can save time, but the owner still needs to check every fact, remove private customer information and make the reply sound like it belongs to the actual situation.

Ask four questions before a review is posted

When AI has touched the process, run the review through four ordinary questions.

If any answer is no or uncertain, stop. Ask for a fresh review in the customer's own words, or do not publish it. One questionable five-star review is worth far less than a review record customers can trust.

1. Did a real customer have this experience? A plausible story is not enough. The service, visit or purchase must have happened.
2. Did the customer supply the facts and opinion? AI should not invent the job, result, staff behaviour, timing or praise. The business should not feed those claims into a prompt either.
3. Did the customer control the final words and post from their own account? A prewritten script is still a script when the customer presses the final button.
4. Does the platform permit both the writing help and the way the review was requested? Check the destination's current rules. Do not assume that an FTC distinction overrides a stricter platform policy.

A safer way to request reviews

Google lets a business create a direct review link or QR code and share it in a thank-you email, chat message or receipt. That removes friction without telling the customer what to say.

A neutral request can be one sentence: “Would you share an honest review of your experience?” Add the direct link, then leave it alone. Do not include a list of keywords, suggested compliments or a five-star instruction. Do not offer a discount, gift, contest entry or future favour.

Make the request part of the normal customer follow-up rather than a rescue campaign when the rating drops. Use the same invitation for eligible customers, including the ones who may have constructive criticism. If the business needs private feedback first, ask everyone the same neutral question and do not route only happy customers toward the public review page.

This approach also produces better evidence. Customers naturally mention the service, problem or constraint that mattered to them. Those details help the next buyer decide whether the business fits. If you are wondering whether review count itself changes ChatGPT recommendations, the evidence supports a more cautious answer.

If a suspicious review appears

Do not paste it into an AI detector and treat the score as a verdict. A detector cannot tell you whether the named customer visited, whether the complaint happened or whether the platform's policy was broken. It can also make polished second-language writing look suspicious when the underlying experience is real.

Start with the business records you are entitled to check. Look for the service, date, location and circumstances described. Note factual conflicts without publishing private information. Then use the platform's reporting process and explain the policy issue you can support, such as no matching transaction, an impossible service claim, a conflict of interest or a pattern of accounts reviewing unrelated businesses with the same text.

If you reply publicly, stay calm. Say that you cannot identify the visit from the details provided and offer a private way to resolve it. Do not accuse the reviewer of using AI unless you have evidence beyond the writing style.

Keep AI on the business side of the line

AI can help an owner organize themes from feedback, draft a first version of a response or spot recurring service questions. It should not become a customer, impersonate one, or supply praise for a customer to post.

The cleanest boundary is easy to explain to staff and agencies: the business may make it convenient to leave an honest review, but the customer owns the experience, the opinion, the words and the account. Check the platform before adding any writing assistance to that process.