Thirty minutes. No pitch. Just a map of what your marketing is missing. Send your brief →
FIELD GUIDE 10
The AI-proof review guide
Customers used to read your reviews. Now a machine reads them and writes a paragraph, and that paragraph is what your customer actually sees.
FIG. 01 · THE LAYER BETWEEN YOU AND THE CUSTOMER
- 82%of consumers read AI-generated review summaries. BrightLocal, Local Consumer Review Survey 2026.
This changes what a review is for. A summariser is not counting stars, it is reading language and looking for repeated themes. Ten reviews that all say "turned up when they said they would" produce a summary that says you are reliable. Ten reviews that say "great job, thanks" produce a summary that says almost nothing.
Vague praise is now a liability
"Great service, highly recommend" was always a slightly weak review. It is now close to worthless, because there is nothing in it for a summariser to extract. The businesses that summarise well have reviews that mention specific, repeated attributes.
| Review says | What the summariser can extract |
|---|---|
| "Great job, thanks" | Nothing. Sentiment only |
| "Turned up on time and cleaned up after" | Punctual, tidy |
| "Quoted 340 and charged 340" | Transparent pricing, no surprises |
| "Came out same day for a burst pipe" | Emergency response, fast |
| "Explained the options without pushing the expensive one" | Honest, consultative |
How to get specific reviews without writing them
You cannot tell a customer what to write. That is both against every platform’s rules and, since October 2024, potentially a violation of the FTC review rule. What you can do is ask a question that makes specificity the natural answer.
Instead of "please leave us a review"
If you have a minute for a review, the thing that helps most is a line about what the job was and how it went. [LINK]
For a business with one clear differentiator
Thanks [FIRST NAME]. If you leave a review, mentioning what you needed and how quickly we got to you is the bit other people find useful. [LINK]
THE LINE YOU MUST NOT CROSS
Asking for detail is fine. Asking for positivity is not, and dictating content certainly is not. "Tell them what the job was" is a legitimate request. "Please mention how great our prices are" is instructing the content of a testimonial, and it is exactly what the FTC rule is aimed at.
Reply in a way the summariser can use
Your replies are part of the corpus. A reply that says "Thanks!" adds nothing. A reply that names the job adds a fact, in your own words, attached to a verified customer experience.
A reply that feeds the summary
Thanks [FIRST NAME]. Same-day callouts on burst pipes are the thing we try hardest to protect capacity for, so glad we got there quickly.
The checklist
- Read your last twenty reviews as a block and write down the themes that repeat. That is roughly your current summary.
- If no theme repeats, your reviews are too vague. Change the ask, not the customers.
- Pick the two attributes you actually want to be known for, and make sure your ask invites them naturally.
- Keep the flow going. Summarisers weight recent reviews, so an old corpus produces a stale summary.
- Reply with specifics, not thanks.
- Read your own AI summary every month. On Google it appears above the reviews. Ask an assistant about your business too.
- Never dictate what a review should say, and never route unhappy customers away from the link.
FIG. 03 · SOURCES, CHECKED 2 AUGUST 2026
Every number above was traced to its primary source before publication. Where a widely repeated figure could not be verified at the source, it was cut rather than repeated.
- VerifiedBrightLocal, Local Consumer Review Survey 2026 1,002 US adults. 82% of consumers read AI-generated review summaries.
- VerifiedFTC, 16 CFR Part 465 Basis for the boundary between requesting detail, which is permitted, and dictating or incentivising content, which is not.
- UnverifiedThe "23% would rely solely on AI summaries" figure Appears in some secondary coverage of the same survey. Not independently confirmed at source on 2 August 2026, so it is not used in this guide.
RANKTHREAD
The summary is the new shopfront
Most businesses are still optimising a star rating that fewer and fewer people look at directly. The thing being read is a paragraph written by a machine from your review text, and that paragraph is shapeable.
RANKTHREAD
Get the next one
We publish a new field guide most weeks. Plain method, sources cited, no pitch. Leave your address and we will send the next one.
