Your star rating matters less than the words under it

Google Maps now sums up your reviews with AI, and chat tools read them too. The words under your stars get repeated. Here is how to ask for better ones.

Caleb Hellinger in front of a white sports car

For about fifteen years the review game had one scoreboard. Get the average up, keep it above 4.5, and push the count higher than the business down the street. Owners printed QR codes, trained front desk staff to ask, and checked the number every Monday.

That scoreboard is losing its job. My view is that the star average now matters less than the sentences people write under it, because the first reader of those sentences is often a machine that turns them into a summary. A 4.8 with two hundred reviews that say "great service!" gives that machine almost nothing to repeat. A 4.6 with sixty reviews that describe what happened gives it a lot.

Google already reads your reviews for the customer

In October 2024 Google added Gemini features to Google Maps, launching first in the U.S. on Android and iOS. One of them is a review summary that sits on a place listing, written by AI from what customers posted. Another lets a person type a question about a place, such as whether it is quiet or has outdoor seating, and get an answer drawn from reviews and other details.

Think about what that does to a person deciding between two clinics or two firms. They used to scan the stars, read the top three reviews and maybe sort by lowest rating. Now a paragraph tells them what reviewers tend to say. If your reviewers tend to say nothing specific, the paragraph is vague, and vague reads as forgettable next to a competitor whose summary mentions short wait times, clear pricing and a front desk that called back the same day.

The same logic carries over when someone asks AI about you in ChatGPT or Claude. Those tools work from text. A star average is one number. The text under it is where the detail lives, and detail is what an answer can quote, paraphrase or compare. I made a related point about third party proof in Proof beats promises, and AI tools know it. Reviews are the most common form of that proof most owners have, and most of it is wasted on one-line praise.

Volume became a trap, and the law agrees

The push for volume did more than produce thin reviews. It pushed a whole industry toward faking them, and regulators noticed.

In August 2024 the Federal Trade Commission finalized a rule on consumer reviews and testimonials that took effect on October 21, 2024. It bans fake reviews, buying reviews that express a particular sentiment, suppressing negative reviews, and buying fake signs of social influence such as followers. It lets the FTC seek civil penalties against people who violate it knowingly. Google's own Maps content policy lists discouraging negative reviews, or selectively asking happy customers for positive ones, as prohibited behavior.

I bring this up because the shortcut most owners reach for, asking only the clients they know loved them, is the one both the platform and the regulator have flagged. The honest version of a review program asks everyone and accepts what comes back. It also turns out to be the version that produces the most useful text, because a mixed set of specific reviews reads as real to a person and gives a summary something to work with.

I would rather show a prospect a 4.6 full of detail than a 4.9 that looks curated. People have learned to distrust a wall of perfect scores, and a review summary will say "some reviewers mention" when the record is mixed, which sounds a lot more believable than a paragraph of uniform praise.

Ask a better question, then stay out of the way

Most review requests are one line: "If you enjoyed working with us, would you leave us a review?" That line invites a one-line answer. It also screens for the happy ones, which is the habit Google's policy warns against.

You cannot write the review for your client, and you should not hand them wording. You can change what you ask and when you ask it. Here is what I would change.

Ask everyone at the same point in the process. Pick a moment, such as the day after a matter closes or a week after a treatment, and send the request to every client who reaches it. A fixed trigger removes the temptation to choose who gets asked.

Ask an open question about the experience. Something like "What was it like working with us, from your first call to today?" invites a story. The client decides what to say. Many will mention the part that mattered to them, and that is the detail a future reader needs.

Ask while the details are fresh. A request sent three months later gets "they were great." A request sent within a few days gets the name of the person who helped, how long things took and what surprised them.

Reply to the reviews you get, including the critical ones. A short, specific reply shows a reader how you handle problems. It also adds your side of the story to the same page, in plain words, without arguing.

None of this guarantees a better summary. The summary is Google's or a chat tool's to write, and I have no way to promise what it will say. What you control is the raw material it reads.

Where this fits in the bigger picture

Reviews are one input among several. A stranger who looks you up also sees your website, any press about you and whatever outside pages describe your business. I wrote about the press side of that in Old press about you still shapes what AI says today. Reviews have one advantage over all of those. They keep arriving every week with no pitch, no editor and no budget, as long as you keep asking.

That makes them the cheapest place to start for law firms, clinics, brands and entrepreneurs who feel invisible when someone asks AI for a recommendation. You already have the clients. You already have the listing. Most owners are collecting the wrong kind of review from them.

Your review audit for this week

Set aside forty minutes.

  1. Open your Google Maps listing on a phone. If a review summary appears, screenshot it with the date. Write down whether it says anything a competitor could not also claim.
  2. Read your last twenty reviews. Mark each one as specific (names a person, a step, a timeframe or a result) or generic. Count the specific ones.
  3. Find the review request you send today. Rewrite it around one open question about the client's experience, and remove any line that tells them what to say.
  4. Choose one fixed trigger point, put it in your process, and send the new request to every client who reaches it from now on.
  5. Reply to your three most recent reviews, starting with the most critical one.

In ninety days, read your last twenty reviews again and redo the count. If the specific ones went up, the summary has more to work with, and so does every stranger who reads your listing before they call.

Want people to find you when they ask AI who to trust?

Keep reading

More from Caleb