7 min read
“Pollute the marketplace and divert business away from honest competitors.” That is how the Federal Trade Commission’s chair described fake reviews when the agency announced its final rule banning fake reviews and testimonials. The rule has been in force since October 21, 2024, and it names AI-generated reviews directly.
Most solo owners read that and think, “Fine, I would never invent a customer.” Good. But the risk in 2026 is quieter than inventing a customer. It is the polite little shortcut: a happy client sends a two-line text, you paste it into ChatGPT, ask it to “make this sound better,” and publish the shiny version on your website with the client’s name under it.
My position is simple and I will defend it for the rest of this article: AI should touch everything around your reviews and nothing inside them. Let it help you ask, sort, reply and learn. Never let it write, rewrite, “polish” or invent the words a customer said about you.
What the rule actually covers (and why “polish” is the trap)
The FTC rule is not aimed at small businesses that sometimes ask for feedback. It targets deception. According to the agency and to coverage such as Search Engine Land’s summary of the rule, it prohibits, among other things:
- Reviews or testimonials that misrepresent who wrote them, including reviews by people who do not exist, which explicitly includes AI-generated fakes.
- Reviews that misrepresent the reviewer’s actual experience with your product or service.
- Buying or selling fake reviews, and paying for reviews that must be positive or negative.
- Insider reviews without disclosure, such as a friend, relative or employee posting without saying they are connected to you.
- Review suppression, meaning threats or tricks used to stop people from posting honest negative reviews.
Now look at the “make it sound better” move through that lens. Your client wrote, “Maria fixed our logo fast, thanks!” Your AI rewrite says, “Maria transformed our entire brand identity in record time, and our sales grew immediately.” That second sentence describes an experience your client never claimed. It is still their name, but it is no longer their review. That is exactly the gap the rule is about, and you will not find a lawyer who tells you “the AI wrote it” is a defense.
I am not a lawyer and this is not legal advice. But you do not need a legal opinion to see the pattern. A testimonial is evidence. Editing evidence with a tool built to make text more persuasive is a bad idea, whatever the tool is called.
The hype version vs the honest version of “AI for reviews”
Search for “AI review tools” and you will find plenty of products that promise more five-star ratings. Some of what they do is genuinely useful. Some of it walks right up to the line. Here is how I separate the two.
Where AI earns its keep
- Asking at the right moment. AI can draft a short, warm review request you send to every customer after a finished job, the same message to everyone, happy or not.
- Replying to reviews. Drafting a thoughtful response to a review is your speech, not the customer’s, so AI help is fair game. We walked through a full system in how to answer every Google review with AI in 15 minutes a week.
- Finding patterns. Paste your last 50 reviews into an assistant and ask what customers praise and complain about most. That is market research, and it is excellent.
- Choosing what to feature. AI can help you pick the three most specific real reviews for your homepage, as long as you quote them word for word.
Where AI turns into a liability
- “Enhancing” a customer’s words before you publish them with their name.
- Writing “sample” testimonials for a new website “until real ones come in.” Placeholder praise that ships is fake praise.
- Only sending review requests to customers the AI predicts are happy. Filtering who gets asked based on expected sentiment is the kind of selective gating regulators and platforms dislike, and Google’s own policies already frown on it.
- Drafting a review “for” a friend to post. Even if they really did buy from you, the words and the missing disclosure are the problem.
“But my customer gave me permission to clean it up”
This is the strongest objection, so let’s take it seriously. Fixing a typo or trimming a long review is common practice, and there is a reasonable middle ground. The test I use is whether the customer would recognize the result as their own words and their own claim.
Practical rules that keep you on the safe side:
- Trim, do not add. You can shorten with an ellipsis. You cannot add a benefit, a number or an adjective the customer did not use.
- Typos only. Fixing “recieve” is fine. Turning “good job” into “exceptional craftsmanship” is not.
- Send it back. If you changed anything beyond spelling, email the customer the exact final version and get a “yes, publish that” in writing.
- Keep the original. Save the raw message or screenshot in a folder. If anyone ever asks, you can show what was said.
Notice what is missing from that list: running the text through an AI writer. Once a model “improves” a testimonial, it is very hard to prove which claims came from the customer and which came from autocomplete.
Why the honest path also wins on marketing
Here is the part that should make this an easy decision rather than a grudging one. Polished testimonials do not even sell well anymore.
Buyers have now read years of AI-generated copy. The overly smooth, adjective-stacked testimonial is exactly what they have learned to skim past. The review that converts is specific and slightly awkward: “She came back on a Saturday to fix the hinge she installed, no charge.” Nobody writes that sentence to sound good, which is why people believe it.
There is also a newer reason. AI assistants increasingly summarize what the web says about a business when customers ask for recommendations. We covered how that can go wrong in our beginner’s guide to AI citations. Real, varied, consistent reviews across platforms are the raw material those systems read. A page of suspiciously uniform praise is weak material for a human reader and for a machine.
A 20-minute review policy you can set up this week
You do not need software for this. You need a short written policy and a habit.
- Write one review request message (AI can help) and send it to every completed customer, not only the happy ones.
- Create a “Reviews” folder where every testimonial you use lives in its original form, with the date and source.
- Add one line to your website’s testimonial section, such as “Reviews shown are quoted from customers, lightly trimmed for length.” If you already have an AI disclosure page, link it there. Our guide to writing an AI disclosure clause and website policy shows how.
- Use AI for replies and analysis only. Write that rule down so future you, tired on a Friday night in December, does not talk yourself out of it.
- Audit your current site once. Look at every testimonial you show today. If you cannot find the original words, take it down or ask the customer to confirm it.
The bottom line for a one-person business
The temptation is understandable. You did great work, the customer said something nice in a hurry, and AI can make it sound like the praise you deserve. Resist it. Your reviews are the one part of your marketing that has to be written by someone other than you, and that includes your AI assistant.
Let AI make you faster at asking, replying and learning. Leave your customers’ words exactly as they wrote them. Which testimonial on your site would you be least comfortable showing the original for? Start your audit there.
Related reading
- How to Answer Every Google Review With AI in 15 Minutes a Week
- How to Tell Clients You Use AI: Write a Disclosure Clause and Website Policy in One Hour
- Is a Fake Site Speaking for Your Business in AI Answers? A Beginner’s Guide to AI Citations
- How to Make AI Write in Your Own Voice: 9 Quick Wins for Solo Business Owners



