Somebody else is writing your description
When somebody asks an assistant to recommend a plumber, the couple of sentences it writes about each business are assembled from whatever it can find, and customer reviews are a large part of that. On a rankings page like this one they are the whole of it: the description we publish is built from reviews and nothing else. Meanwhile the paragraph you wrote about yourself, probably years ago, sits on your own website where almost nobody reads it.
So a review has a second job it did not used to have. A five star rating on its own is a vote, and it gets counted. A five star review saying "replaced our hot water cylinder in Papakura on a Saturday, quoted before starting, no mess left" is material somebody can quote. It says what you do, where you do it, how you work, and what to expect.
The difference shows up when the question carries a condition. Darren Shaw's example is a customer asking for a plumber and adding that they have just had new floors laid and do not want them wrecked. Answering that needs a business whose reviews mention leaving a place tidy. If yours do not say it, you are not a candidate, however careful you actually are. Every attribute your reviews name is a job you can be picked for.
Brad Weatherall, formerly of Google, draws the line between the two sides of it: your business profile is you telling customers about yourself, while reviews and customer photos are other people telling the world about you. The second carries more weight because you did not write it.
Both he and others in the field expect AI systems to lean harder on third-party material over time. That is an expectation, not a measurement, and worth holding loosely.
We see the effect from the inside, since we write those summaries ourselves. Each of the roughly 4,400 business descriptions published across our pages is written from up to 25 of that business's reviews, and up to 50 on a paid listing. Plenty of businesses fill that comfortably. Others have fewer than ten reviews carrying any usable text, and that is the ceiling on what we can say about them.
A deep set gives us enough to cover the range of work a business does. A thin one gives us a narrow paragraph that can make a broad business look like a one-trick operation. Several beta testers told us their summaries read as reductive, and they were right. The reviews were the constraint.
How to ask so you get something usable
Most review requests ask for a review. The better ones ask for a description of what happened, which is a small change to the wording and a large change to what comes back. Instead of "could you leave us a review", try "could you leave a review, and say what the job was and how it turned out". Ask everyone, not only the ones you expect to be kind, because selectively asking happy customers is against Google's policy and is the practice most likely to get a profile flagged.
You cannot tell a customer what to write. Google's policies do not allow steering either the content or the rating, and they mean it. Asking somebody to describe their own experience is a different thing from telling them what to say about it.
Four other things shape what you end up with:
- Who you ask. If every review you hold came from an emergency callout, then in every description written about you that is what you do, even when half your work is planned installations. Ask across the range of jobs, and lean towards the clients whose work you would like more of.
- When you ask. A cluster of reviews from one busy month describes one busy month, where reviews arriving steadily across a year describe a business that does this consistently.
- What you ask for. Send identical wording to everyone for a year and you get a year of reviews about the same thing. Keep several versions of the request going and rotate them, each prompting for a different specific: what the job was, how quickly you got there, how the place was left, whether the quote held.
- Where you ask. Reviews on sites other than Google get read as well, and which ones count varies by trade. Search a competitor's name plus the word reviews and see which sites Google returns for them.
Telling us what to lead with
Everything above applies wherever your reviews get read. On our own pages there is one more lever, and it is the clearest thing a Top5.nz subscription buys.
Subscribers get a field in their dashboard headed What should we highlight?. You put in up to three things, and we use them to decide what the description of your business leads with. Paid listings also get a longer write-up, 60 to 80 words against 45 to 65, which is more room to cover the range of work you do rather than the one job your reviews mention most.
How it works:
- It changes the order, not the content. Everything published still comes out of your customer reviews. An item your customers do not write about is ignored rather than worked in, and we know because we tried it: "commercial contracts" on a business whose reviews never mentioned commercial work never once appeared, across five runs.
- It works best on the things you are not already known for. We lead with whatever your reviews talk about most anyway, so the field makes least difference there and most difference to the work sitting second or third.
- Concrete beats abstract. "Our technicians by name" is the kind of thing that works. Values and adjectives do not, because a description built from evidence has nothing to attach them to.
- Your wording never reaches the page. The field points at reviews you already have, it is not copy. In the instructions we give the model it sits after every editorial rule, so it cannot loosen one, and an adversarial test of superlatives and swipes at competitors was stripped in full across three runs. That is why nothing you type there needs approving: there is nothing to approve.
- It has no effect whatsoever on your ranking position. That is set by the published method, and no business can pay to move.
Your summary is rewritten at the next refresh of your category, usually within a month, and we email you the new wording when it changes.
Find out what yours currently say
Ask an AI assistant to summarise your business, or to recommend someone in your trade and city, and read what it gives you. That is about as close as you will get to seeing the description your customers have written on your behalf.
Read it for what is missing rather than what is wrong. There is usually a service you offer that never comes up, or a strength you are known for locally that no reviewer has ever put into words. That list of gaps is what to steer the next few months of asking towards.
Keep replying to reviews as well, including the awkward ones, since the replies form part of the same body of text being read. Write them for the person, not the machine: you can tidy a reply up, but stuffing it with the services you want mentioned reads exactly as badly as it sounds. Our guide to responding to negative reviews covers how to do that without making things worse.