There is a paragraph on most of your detail pages that you did not write, cannot edit, and that a meaningful share of mobile shoppers read before they read your bullets. It starts with the words “Customers say.” Amazon generates it from the text of your reviews, refreshes it as reviews accumulate, and hangs a row of tappable attribute tags under it, some of them green and some of them not.
After managing hundreds of brands on Amazon, we have watched sellers spend five figures rebuilding an image stack while that paragraph quietly told every shopper the product “runs small” and the “instructions are unclear.” The stack was fine. The paragraph was the conversion problem. And because nobody in the business owns it, nobody was reading it.
This post is about treating Amazon’s “Customers say” review highlights as what they are: a creative asset with a 60-to-90-day lag, authored by your buyers, edited by a model, and shaped almost entirely by what your listing promised before the box arrived.
What the feature actually is
Amazon introduced AI-generated review highlights in 2023 and has expanded them since. The mechanics, per Amazon’s own description: the model reads the text of customer reviews, extracts the sentiments mentioned most often, and produces a short paragraph summarizing shared positive, neutral, and negative opinions about the product and its features. Under the paragraph sit tappable attributes (things like “Value,” “Comfort,” “Size,” “Instructions”) that filter the reviews to the ones mentioning that theme.
Two things about the mechanics matter more than the rest.
It is frequency-weighted, not rating-weighted. A product at 4.6 stars whose three-star reviews all say the same sentence will see that sentence surface. Sellers on the forums complain about exactly this, and they are right that it feels unfair. They are wrong that it is arbitrary. The model is doing what a frequency model does: reporting the thing said most often. Which means the paragraph is a compressed version of your most repeated expectation gap.
It sits high on mobile. On the app, the highlights block frequently renders above the individual reviews and, depending on category and Amazon’s ongoing layout experiments, near enough to the buy box that a shopper who never scrolls into the full review list still sees it. Confirm placement on your own top ASINs in the app rather than off any blog, including this one. Amazon moves detail-page modules by category and device without announcing anything.
Why this is a creative problem and not a review problem
Sellers treat the paragraph as a reviews issue and route it to whoever handles customer service. That is the wrong department.
Read what actually appears in negative highlights across categories: “smaller than expected,” “color was different,” “runs large,” “didn’t fit my,” “instructions were confusing,” “felt cheaper than it looked.” Every one of those is a sentence about the distance between what the page showed and what the box contained. Not one of them is about the product being bad. They are about the product being different.
That distance is set by your creative. The hero image implied a size. The lifestyle shot was graded for appeal rather than accuracy. The compatibility statement was in a bullet nobody read instead of in a frame at position two. The shopper bought the product in the picture and received the product in the box, and the gap between them got typed into a review, and the model counted how many times it was typed.
So when the “Customers say” block reports that customers “mention the product is smaller than expected,” it is not reporting on your product. It is reporting on your slot-two image. And it is doing so in the one location on the page where a first-time shopper is most likely to believe it, because it is visibly not written by you.
The lag that makes this expensive
Here is the sequence on a real-shaped account. A $45 kitchen SKU doing roughly $60K a month launches with a beauty-shot-heavy stack in March. By May, “smaller than expected” is the most common phrase in the three-star reviews. By June, it is in the highlights paragraph. By July, conversion has drifted a point and a half and the brand is looking at ad efficiency because that is where the reporting lives.
The creative decision that caused it was made in March. The paragraph reporting it appeared in June. The conversion effect became visible in July. And the fix, if the brand makes it in August, will not change the paragraph until enough new reviews accumulate to shift the frequency count, which on a SKU at that velocity is another 60 to 90 days.
The “Customers say” paragraph is the slowest feedback loop on the detail page, and it is the one with the longest tail. A bad hero costs you clicks this week. A bad highlights paragraph costs you conversion every week until the review distribution changes, and it sits there through Q4 while you are paying peak CPCs to send traffic to it.
Going into the ten highest-value weeks of the year, a negative theme in that block is not a reviews problem to fix in January. It is a creative problem to fix in the next three weeks, because that is the last window where a listing change can start moving the review distribution before peak traffic reads it.
The 20-minute audit
Pull up your top ten revenue ASINs in the Amazon app. Not desktop, not Seller Central. The app, as a customer.
That audit produces a ranked shot list. It is not a design exercise. It is the buyers telling you, through a model, which questions the page failed to answer.
What actually moves the paragraph
You cannot edit it, request changes to it, or appeal it in any meaningful sense. Sellers on the forums have tried. What you can do is change the inputs, and the inputs are the reviews of the next 90 days, which are shaped by the expectations the page sets today.
The scale reference at slot two. The single most common negative highlight in physical goods is some version of “smaller than expected.” A hand, a countertop, a doorway, a person, placed early enough in the stack to correct the impression the hero set. This is the highest-return creative change available on most listings and it requires a phone, not a photoshoot.
Color accuracy over color appeal. If “color was different” is in your highlights, the saturation that won you the click is producing the review. Shoot with a grey card, grade toward accuracy, and where color genuinely varies, say so in a frame.
The compatibility answer in type heavy enough to survive a phone. “Didn’t fit my [thing]” in the highlights means the compatibility statement is in a bullet or a spec table and the shopper guessed. Move it to a frame. State it plainly. Use a weight that renders.
The what-arrives frame. “I thought the stand was included” and its cousins come from a stack that showed the product styled rather than boxed. One two-panel image of the closed carton and its contents answers a question that otherwise gets answered by a review.
The honest fine print. If a limitation is real, put it on the page. A shopper who reads “not for induction cooktops” and buys anyway does not write “doesn’t work on induction” in a review. A shopper who was never told does.
None of those are rebuilds. Four of the five are shippable inside three weeks from files and phones you already have. And every one of them changes what the next 200 reviews say, which changes what the paragraph says by November.
The measurement
Screenshot the “Customers say” block on each top-ten ASIN this week with the date in the filename. Do it again monthly. Amazon does not archive prior versions of the paragraph, and the block is not exported in any report, so if you do not capture it you cannot prove it changed.
Grade the creative fix against three things, in this order:
- Return-reason comments on the same theme, 60 days out. This moves first because it does not depend on review velocity.
- The attribute tags under the highlights, 90 days out. A theme that was a negative tag and is now absent or neutral is the fix landing.
- Conversion rate, 90 to 120 days out, with the honest caveat that Q4 traffic mix will contaminate any read that spans October. Compare like periods.
Do not judge this inside 30 days. The paragraph is downstream of reviews, reviews are downstream of deliveries, and deliveries are downstream of the page. Three lags, stacked. Anyone who tells you a listing change moved the highlights in a fortnight is describing a coincidence.
FAQ
Can I get Amazon to remove or change the “Customers say” summary?
Not in any reliable way. It is generated from review text and updates as reviews accumulate. Sellers have opened cases; the standard response is that the summary reflects customer reviews. The only durable lever is the composition of future reviews, which your listing sets.
Why does it highlight a negative when my rating is 4.6?
Because it is frequency-weighted. A theme mentioned in 15% of reviews will surface even if 85% of buyers are happy. This feels unfair and is actually useful: it is the most-repeated expectation gap on the listing, surfaced for free.
Does the highlights paragraph affect ranking?
Not directly in any way Amazon has published, and anyone quoting a weight invented it. It affects conversion, and conversion is a ranking input. Treat it as a CVR lever, not an SEO one.
Should I respond to the reviews it surfaces?
Seller responses to reviews have limited visibility and do not change the summary. Spend the hour on the frame that prevents the next review instead of on the reply to the last one.
Is it too late to affect the paragraph before Q4?
For a high-velocity SKU, a creative change live by mid-September will start shifting the review distribution by November. For a low-velocity SKU, probably not. Make the change anyway; the reviews it prevents in December are the ones that would have sat in the paragraph through Q1.
The paragraph is going to be on your page either way. The only decision is whether the buyers who write it were shown the product they received.
If you’re looking for a team that manages every lever — creative, advertising, and operations — Velocity Sellers works with brands doing $100K+/month on Amazon. Contact us for a free account audit.
