Blog/How-To
HOW-TO

How to Ask Clients for Reviews That AI Can Find and Trust

Five round rating markers of dark smoked glass standing in a row in a deep blue void, the first four glowing warm amber from within and the fifth left dark

Key Takeaways

  • Send them to a public platform, not your inbox. A glowing message in your messages app is worth nothing here. A review has to be published somewhere a machine can reach before it can be repeated.
  • Ask questions, not for a review. "Could you leave a review" produces one vague sentence. Three specific questions produce the details that make a review quotable.
  • Get the starting point into the text. A review that says where someone began is far more useful than one that says you are great, because it matches the question a buyer is actually asking.
  • Put a transcript under every video testimonial. Spoken praise is invisible as text. The transcript costs a few minutes and turns an unreadable asset into a readable one.
  • Never trade anything for a positive review. An incentive can never be conditioned on the review being favourable, many platforms ban incentivized reviews outright, and a run of similar-sounding reviews reads as fake to people and machines alike.

Most coaches have better proof than they can show. It arrives as messages, voice notes, and the thing a client says on the way out of a session, and almost none of it ever lands anywhere a machine can read.

Meanwhile the review request that does get sent produces one polite line about how lovely you are, which is pleasant and completely unusable.

This is about fixing both ends: where a review has to live, and how to ask so the answer is worth quoting.

How do you get reviews that AI can actually use?

Ask three specific questions instead of asking for a review, then send people to a public platform to post the answers. The specifics give an assistant something to quote, and the public platform is what makes it readable at all.

Both halves are required. A brilliantly detailed review sitting in your email is invisible. A public review that says "highly recommend" is visible and says nothing. The reviews that get repeated in answers do both jobs at once.

Nothing here asks you to manufacture anything. Every specific in a good review is a fact your client already knows and simply was not asked for.

Why most reviews are invisible

Because they are not text, or not public, or not attached to any detail worth repeating. Those are three separate failures and most coaching businesses have all three at once.

Not text: a screenshot of a message is an image, and the words inside it are not words as far as an automated reader is concerned. The same is true of praise inside a video with no transcript, and of a testimonial slider whose contents load after the page has been read. The full version of that problem is in making client results legible to AI.

Not public: the best sentence anyone has ever written about your work is worth nothing here if it lives in a direct message. Not detailed: "great coach, highly recommend" matches no buyer question, because nobody asks an assistant to find them a coach who is great.

What a quotable review contains

Five things, and none of them require your client to write well. They require you to ask better questions, because people answer the question they were asked.

WHAT SURVIVES BEING QUOTEDILLUSTRATIVE SPECIMENS
THE USUAL REVIEW One usable phrase in three lines. An assistant summarizing this has nothing specific to hand back, so it hands back somebody else.
THE ONE THAT TRAVELS Five liftable details in the same three lines. Each one matches a question a buyer actually types.
  1. Who they were at the start Complete beginner, returning after injury, training for a first event.
  2. What actually changed The specific result, in the words they would use themselves.
  3. Over how long A timeframe turns a claim into something checkable.
  4. Which service they bought Small group, one to one, the twelve-week programme.
  5. Where they are The town or the online context, so the review matches a local question.
answerhalo.com
Two specimens drawn as bars, not quotes. Nothing in this figure is a real or invented review. The lit segments mark the five details listed beneath, all of which are written out in the text.

In text, because the figure holds no claim of its own. The five details are who they were at the start, what actually changed, over how long, which service they bought, and where they are. A review carrying all five answers a buyer's real question. A review carrying none of them answers nothing.

This is the same property that makes any content quotable, which is covered more generally in why AI lifts tables and lists: a unit that survives being separated from everything around it. A sentence that says "I went from never having trained to deadlifting my bodyweight in five months with the small group sessions" survives on its own. "She's brilliant" does not.

Where a review has to live to be read

The platform matters as much as the words. Here is where reviews usually end up, and what each location is actually worth when an assistant is assembling an answer about you.

Where a review lives, and what it is worth to an assistant reading about you
Where it lives Readable as text? Independent of you? What it is good for
Google Business Profile Yes Yes The single highest-value place to send people
A well-known industry platform Yes Yes Niche-specific questions, second priority
Plain text on your own site Yes No Detail and context you control, still quoted
Testimonial slider that loads after the page Often not No Reassuring visitors, not being read
Video testimonial with no transcript No No Warmth on the page, invisible as evidence
Screenshot of a message No No Nothing, unless you also write it out
Direct message or email to you No No Your morale, and permission to ask for more
A story or post that expires No No The day it runs, and nothing after

The bottom half of that table is where most coaching proof currently sits. That is the uncomfortable finding, and it is also the cheap win: the evidence already exists, and moving it costs an afternoon of copying with permission rather than a year of new results.

If you want the technical form of a review that a search engine reads without guessing, Google documents its review snippet markup and the underlying vocabulary is the schema.org Review type. Useful later. Text on the page first.

The ask that gets specifics

Stop asking for a review. Ask three questions, then offer to turn the answers into the review, and let them post it in their own words.

  1. Ask at the moment of the result, not at the end of the engagement. The week someone hits a milestone is the week they can describe it. Three months later they remember that it went well.
  2. Send three questions, not a request. Where were you when we started? What changed, and roughly how long did it take? What would you say to someone deciding whether to book?
  3. Send the link with the questions. One link, one platform, and the questions in the same message. Two links halves your response rate.
  4. Offer to tidy, never to write. You can offer to fix typos. You cannot write it for them, and a review you drafted is not a review.
  5. Transcribe anything spoken. If a client sends a voice note or films something, publish the words as text with their permission. Spoken praise is invisible until it is written down.

Two things to avoid, and both matter more than they used to. Do not offer anything in exchange for a positive review: the FTC's guide to soliciting and paying for online reviews says an incentive must never be conditioned on the review being positive and should be disclosed when it exists, and it notes that many platforms prohibit incentivized reviews outright. Do not post client results without permission, however proud you are.

Then check whether any of it is landing. Ask an assistant what people say about your business and see whether the answer sounds like your actual clients. The fifteen-minute test covers the wider version of that check.

Our free check asks twenty-five buyer questions for you at $0 and emails a plain-English readout within the hour. It runs as soon as you ask for it.

Questions we hear the most

How do I ask clients for reviews that AI can find and trust?

Ask them three specific questions instead of asking for a review, then send them to a public platform to post the answers. The specifics give an assistant something to quote, and the public platform is what makes it readable in the first place.

Do AI assistants actually read reviews?

They read what is published publicly and summarize it, which is why review text now matters more than star counts. A five-star average with no words behind it gives an assistant nothing to say about you.

Where should I send clients to leave a review?

Your Google Business Profile first, then one industry platform that is relevant to your niche. Two well-populated public sources beat five thin ones, and reviews on platforms nobody indexes may as well be private.

Are testimonials on my own website worth anything?

Yes, if they are plain text on the page rather than images or a slider that loads after the page does. They carry less weight than an independent platform because you control them, but they are still readable and still quoted.

Can I offer a discount in exchange for a review?

Not for a positive one. Consumer protection guidance says an incentive must never be conditioned on the review being favourable and should be disclosed where it exists, and many review platforms ban incentivized reviews outright regardless.

How many reviews do I actually need?

Fewer than you think, if they are specific. Ten detailed reviews that name a starting point, a result, and a timeframe do more for you than a hundred that say great coach, highly recommend.

How do I find out whether my reviews are being used in AI answers?

Ask an assistant what people say about your business and read what comes back. If the summary is vague or wrong, the reviews it can reach are either thin, stale, or somewhere it cannot read.

What does the AnswerHalo free check include?

It costs $0. We ask ChatGPT 25 real buyer questions from your niche, then email a plain-English readout within the hour: how many answers named you, who was named instead, and the first fix. It runs as soon as you ask for it.

SEE WHERE YOU STAND

Your buyers are already asking. Find out what AI tells them.

The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.

Get my free check →