Review platforms and the schema that makes them count

What reviews actually do for search

Start by separating three effects that get bundled together in every review app’s pitch.

Appearance in the result. Valid rating markup on a product page can put a star rating and a count into the search result. That is a click-through effect. The same position, with stars, gets more clicks than the same position without, which is worth having and is not a ranking improvement. Nothing about the star changes where you sit.

Content on the page. This one is underrated. Reviews add the words customers use, including the fit complaints, the comparisons and the use cases nobody on the marketing team would think to write. That is real content, it is unique to your page, and it is the sort of specific, attributable language answer engines quote. It is also the reason a product page with forty reviews behaves differently in search from the identical page with none. The page is different.

Ranking, directly. No. There is no review count input you can turn up. If the third-party comparisons you read imply otherwise, look at what is being compared: pages with many reviews tend to be pages for popular products on well-linked sites, which is the variable doing the work.

The business case is therefore conversion first, click-through second, content third, and ranking not at all. Which is fine: conversion is the better prize anyway.

The rule that stops most brands doing what they want

Almost every brand, on discovering review markup, wants the stars next to their homepage. Google’s review snippet documentation closes that off in one sentence: "If the entity that’s being reviewed controls the reviews about itself, their pages that use LocalBusiness or any other type of Organization structured data are ineligible for star review feature." Organization reviews are acceptable when you are reviewing *other* organizations (a directory, a comparison site), not yourself. The guidance goes further on curation: "Don’t rely on human editors to create, curate, or compile ratings information for local businesses."

The restriction is not arbitrary. Self-collected, self-published, self-selected ratings are not evidence. Marking them up as if they were invites a manual action that can remove every rich result on the domain: not just the stars you were reaching for. The downside is asymmetric and the upside was a decoration.

So put the company-level number on the page as visible text instead. "Over a hundred five-star reviews", stated plainly and attributably, does the persuasion job for the human reading it and carries no markup risk at all. We follow that rule on this site: no aggregateRating for the organization, review counts as copy.

Product-level ratings are a different matter and are the sanctioned route: a product review is a review of a thing you sell, not of you, and Google supports Product in the review snippet types. That is where the markup effort should go.

Two copies of the truth: the Shopify-specific fault

We find this bug on most stores that have a review app installed, and it is invisible unless you look at the rendered source.

Shopify themes can emit product structured data themselves. The structured_data Liquid filter “converts an object into a schema.org structured data format”, outputting a product as Product if it has no variants and as ProductGroup if it has one or more. Review apps also emit product structured data, because emitting it is one of their selling points. Install one on the other and the page ships two Product nodes: one with the rating, one without, or worse, two with different ratings because one is cached.

What a crawler does with that is unpredictable and none of the outcomes are the one you wanted. The fix is a decision, not a plugin: one component owns product markup on this store. Usually it should be the theme, because the theme knows about variants, pricing, availability and the ProductGroup distinction, and can read the rating from a metafield the app writes. Then turn the app’s markup off. Most have a setting, and the ones that do not are telling you something.

While you are in there, check the smaller requirements, because they are the ones that silently disqualify a result. AggregateRating needs a ratingValue and at least one of ratingCount or reviewCount. A Review needs an author, the item reviewed with a name, and a reviewRating.ratingValue. And the author has to be a plausible name for a person or team. Google’s product documentation explicitly rejects things like "50% off on Black Friday" in the reviewer name field, which exists as guidance because somebody kept doing it.

Choosing a platform, and why the cheap answer usually wins

Review platforms price from free to several hundred a month, and the expensive tiers sell features that are mostly not the reason you are buying.

Four questions decide it, and "does it produce rich snippets" is not one of them, because your theme can do that and should.

• Can you export every review, with its author, date, rating and product, in a format you could import elsewhere? If not, you are renting your social proof. This is the question that should end most shortlists.
• Does it collect from verified purchasers automatically, tied to real orders? Ratings have to come from customers rather than from you, so the collection mechanism is the product.
• What does it add to the page at runtime? A review widget is frequently the heaviest third-party script on a product page. Ask whether it renders server-side, whether it can be deferred, and what happens to the reviews if the script fails.
• Does it write ratings somewhere your theme can read them, a metafield, so that display and markup can live in your theme rather than in their iframe? This is the difference between reviews you own and reviews you embed.
• And one negative: do not pay extra for a tier whose headline benefit is "Google Seller Ratings" or site-wide stars. Read the rule in the section above before you buy it.

Where the reviews go on the page

Collection and markup are the technical half. The half that changes revenue is placement, and it is mostly free.

The rating summary belongs near the title, above the fold, as a jump link. That is the element shoppers scan for. The reviews themselves belong far enough down that they do not compete with the buy button, and they need to be filterable by rating, because the first thing a careful shopper does is read the two-star reviews. Hiding those is both dishonest and counter-productive: a product with nothing but five-star reviews reads as curated, and it is.

The one thing to build is the ability to lift a specific review into a page. A single well-chosen quote, on a collection page or a landing page, next to the product it is about, does more than a widget of forty. That means blocks a merchandiser can place, which is what a section library is for, and reviews are one of the better arguments for having one.

If reviews are thin, say so rather than dressing it up. Twelve honest reviews displayed plainly beat two hundred that a shopper suspects. And if you are running an incentive to collect them, take advice on how it must be disclosed. Incentivized and fabricated reviews are a consumer-protection issue in the UK, not only a search one.

What we would do, in order

Half a day of checking, then one decision.

• View source on three product pages and count the Product nodes. More than one is the problem to fix first.
• Decide which component owns product markup, normally the theme, and switch the other off.
• Validate a product page and confirm aggregateRating has both a value and a count, and that ratings are attached at the level your theme emits (Product or ProductGroup).
• Remove any aggregateRating attached to your Organization markup, and put the review count on the page as text.
• Check what the review widget costs in page weight, and whether it can render without blocking.
• Only then look at collecting more reviews, and measure conversion on the product pages, not rankings.

Questions this raises

Read next

• Shopify section libraries for content-heavy brands
• The product page audit: twenty things worth checking
• What AI assistants say about your brand, and why