Reviews

The reviews AI actually reads (and why star count is not the signal)

A five star review that says great is worth nothing to a machine. Here is what a quotable review looks like.

September 8, 2026 · 5 minute read · By KKV Labs LLC

Short answer

Assistants read the text of reviews, not only the star average. A review that names the service, the town and what actually happened gives an answer something to quote. Five stars and the word great gives it nothing. Recency matters as well, because a page of reviews from three years ago reads as a business that may have changed since. The way to get useful reviews is to ask soon, ask specifically, and never buy them or filter who gets asked.

Why the star count is not the signal

Almost every established local business sits somewhere between four and five stars. An average inside that band tells a machine nothing that separates you from the four competitors next to you. What separates you is the content: which services get mentioned, in which town, and with what result.

Think of it from the answer's side. If someone asks who is good for a specific service in a specific town, the answer needs evidence tying you to that service in that town. Star averages carry none of that. Sentences do.

What a quotable review looks like

The reviews that get used tend to have three parts, in any order.

  • What they came in for, named plainly. The treatment, the repair, the case type.
  • Where. The town, or the neighborhood, said in passing.
  • What actually happened. How long it took, what it cost, whether the problem was solved, what surprised them.

You cannot script that, and you should not try. But you can ask a question that makes those three parts the natural thing to write, which is most of the job.

How to ask so you get those words

  • Ask soon. The day of the visit or the day after, while the detail is fresh.
  • Ask everyone, not just the customers you expect to be happy. Filtering is against the rules on the major platforms and it shows in the pattern of your reviews.
  • Ask a question rather than begging for stars. What did you come in for, and how did it go, gets you a paragraph. Please leave us five stars gets you a number.
  • Send one link straight to the review form. Every extra tap loses people.
  • Ask by the channel they already use with you, whether that is text, email or a card handed over at the desk.
  • Ask again once, politely, if nothing arrives. Once, not four times.

Never offer anything in exchange. Discounts, prize draws and free add-ons for reviews are against the terms of the major platforms, and they produce exactly the empty reviews that are worthless here anyway.

Where the reviews need to be

  • Google Business Profile first. It feeds Google AI Mode directly.
  • Then the one or two review sites that already come up when you search your service and city. For most trades that is a small and obvious list.
  • Then the industry specific site for your field, if one exists and people in your town use it.
  • On your own site as visible text, if you display them at all. A review widget that pulls them in with script may leave nothing for a machine to read.

Spreading a handful of reviews thinly across ten platforms helps nobody. Pick the two or three that appear in your own search results and concentrate there.

Responding, and what it signals

Reply to reviews, including the bad ones. A reply is more text on the record, in your words, and it is the one place you can name the service and the town accurately without asking anyone else to do it. Keep replies short, specific and calm. Do not argue, and never publish anything about a customer's treatment, case or personal circumstances in a reply.

A negative review answered well reads better than no negative reviews at all. A perfect record with no responses reads like a page nobody is minding.

What to skip

  • Buying reviews. It violates platform terms and consumer protection law in many places, and it is detectable.
  • Writing them yourself, or asking staff and family to.
  • Review gating: surveying people first and only sending the review link to the happy ones.
  • Incentives of any kind.
  • Bulk uploading a hundred old testimonials to your website in one afternoon and calling it social proof.
  • Obsessing over the average. Aim for a steady flow of specific, recent, real reviews and the average takes care of itself.

Do this first

What to do this week.

Eight things, in order. None of them needs a developer or a budget, and every one of them is checkable by you.

  1. 01Pick the two or three review platforms that show up when you search your service and city.
  2. 02Write one short ask that requests a sentence about what they came in for.
  3. 03Create a direct link to your Google review form and save it where you can send it in seconds.
  4. 04Ask every customer from the last two weeks, not only the happy ones.
  5. 05Reply to every review from the last three months.
  6. 06Delete any incentive language from your existing request templates.
  7. 07Check that your website shows review text as real text, not only inside a script widget.
  8. 08Put the ask into your normal closing routine so it happens without you remembering.

Questions

The questions owners ask us about this.

How many reviews do I need?
There is no threshold anyone can honestly quote you. What matters more is that they are recent, specific and spread across the services you actually want to be named for. A steady trickle every month beats a burst of fifty followed by two silent years.
Do bad reviews stop AI naming me?
A few negative reviews among many positive ones are normal and read as normal. What reads badly is a pattern: the same complaint repeated, or a long run of recent unhappy customers with no replies from you. Answer them, fix what caused them, and keep asking everyone else.
Can I ask customers to mention a specific service in their review?
You can ask a question that naturally leads there, such as what did you come in for and how did it go. You should not dictate the wording or hand people a script. A review has to be the customer's own words, and a set of identical reviews is obvious to everyone including the platform.
Should I put reviews on my own website?
Yes, as visible text, and only real ones. If you use a widget, check that the review text appears in the page source rather than being pulled in afterwards by script, otherwise there may be nothing there for a machine to read.
Does replying to reviews actually matter?
It adds text in your own words that names the service and the place, and it shows a business that is being run. Keep it short and never disclose anything personal about the customer, which matters especially in health and legal work.

Or find out what AI says about you before your competitor does.

Free 10-prompt check with screenshots, for your service and your city. No call required to get it.

All nine guidesWhat the free check includes

Questions? hi@kkvlabs.com