Review schema and review snippets

01What is review schema?

Review schema is the schema.org Review type, plus the Rating it carries, written into a page as JSON-LD, Microdata or RDFa. It tells a search engine that the page contains a review, names the thing being reviewed, gives the score and says who wrote it.

A review snippet is the search result that can follow: a row of stars and a score under the page title. The markup is the input; the snippet is an optional output that Google decides on page by page and query by query.

Google describes a review snippet as a short excerpt of a review or a rating, and says that when it finds valid reviews or ratings markup it may show a rich snippet with stars and summary information. The operative word is may. The review snippet documentation lists the item types that qualify and the guidelines that apply, and the rest of this guide works through both.

schema.org defines the Review type as a review of an item, for example of a restaurant, movie or store. The vocabulary lets you review almost anything. Google's support is much narrower, and that gap is where most missing stars come from. If markup itself is new to you, start with what structured data is and why machines read it, then come back here.

02Review or AggregateRating: which one do you need?

The two types answer different questions. Review is one verdict: one author, one score, usually one piece of text. AggregateRating is a summary of many verdicts: an average score and a count of how many ratings or reviews produced it.

Google's review snippet guidelines add one rule that ties them together: if a page includes several individual reviews, it should also include an aggregate rating of those reviews. Ratings collected from your own visitors have their own guide, AggregateRating schema for user reviews, so this page concentrates on single reviews and on the rules that apply to both types.

On a WordPress site the split is usually easy to see. A review box written by the site's editor needs Review: one author and one score. A star widget that collects votes from visitors needs AggregateRating. Many review plugins print both, and a common mistake is an AggregateRating with a ratingCount of 1 built from the editor's own score, which presents one verdict as if it were many.

03Which item types can show review stars?

Google lists the item types that can carry review snippets. As of November 2026 the documented features are Book, Course list, Event, Local business, Movie, Product, Recipe and Software App. Google also supports reviews for CreativeWorkSeason, CreativeWorkSeries, Episode, Game, MediaObject, MusicPlaylist, MusicRecording and Organization, together with their subtypes.

Local business and Organization carry a condition: they qualify only on sites that capture reviews about other businesses or organizations. That condition is the self-serving rule, covered in its own section below.

Anything outside the list can still be valid schema.org, and a generic validator may accept it, but it will not produce stars in Google. A review of a person, or of an abstract service with no supported parent type, is the common example. If the thing you review is something people buy, Product schema is usually the right parent, and the review nests inside it.

Review snippets are one of several features covered in our guide to schema types that still earn results, which sets out what each type can and cannot do today. The other types there follow the same pattern: a supported item, required properties and content readers can see.

04Required and recommended properties

For a single review, Google requires an author, a reviewRating with a ratingValue, and the name of the item reviewed. If the review is not nested inside the item's own markup, it also needs itemReviewed. The author must be a valid name shorter than 100 characters; Google's documentation gives "50% off until Saturday" as an example of a name that does not count.

The scale matters more than it looks. Google reads a bare number on a five-point scale, so a score of 8 out of 10 needs bestRating set to 10, or it is read as 8 on a scale that tops out at 5. Fractions such as 6/10 and percentages carry their own scale.

Decimals need a dot, not a comma, in the markup. In Microdata and RDFa you can show a comma to readers and put the dotted value in a content attribute; in JSON-LD the number is simply written with a dot.

05What changed in 2019: the self-serving review rule

On 16 September 2019 Google announced, in its Search Central post on review rich results, that it would limit review snippets to a fixed list of schema types and stop showing them for self-serving reviews. It called a review self-serving when a review about entity A is placed on entity A's own website, whether directly in the markup or through an embedded third-party widget.

The rule covers LocalBusiness and Organization and their subtypes, and the same post confirms it applies to AggregateRating as well as Review. A plumber that marks up testimonials on its own home page will not get stars for them. A site that reviews plumbers it does not own still can.

The post also answers the two questions owners asked first. You do not need to remove self-serving markup, and having it will not by itself lead to a manual action; Google simply stops showing stars for it. The change did not touch Business Profile, because it applied only to organic search.

The current review snippet guidelines keep the rule and add another for local businesses and organizations: ratings must come directly from users, not from editors who create, curate or compile them. That is why an editor's verdict on a restaurant does not earn stars the way an editor's verdict on a book can.

06How to write an editor review in JSON-LD

The example below is an editor's review of a fictional book. The review is the top-level item, so it names the book in itemReviewed. The score uses a ten-point scale, so bestRating and worstRating are explicit. Every value is an example.

1{2  "@context": "https://schema.org",3  "@type": "Review",4  "itemReviewed": {15    "@type": "Book",6    "name": "Example Book Title",7    "author": {8      "@type": "Person",9      "name": "Example Novelist"10    }11  },12  "reviewRating": {13    "@type": "Rating",14    "ratingValue": 8,15    "bestRating": 10,216    "worstRating": 117  },18  "author": {19    "@type": "Person",20    "name": "Example Reviewer"321  },22  "publisher": {23    "@type": "Organization",24    "name": "Example Review Site"25  },26  "datePublished": "2026-11-01",27  "reviewBody": "A clear, well-paced example of the genre with a weak final chapter."28}

If the page is mainly about the book and the review is one part of it, flip the structure. Mark up the Book and nest the review in its review property, then drop itemReviewed, because Google assumes the parent item is the one being reviewed.

Either way, the reader must be able to see the review text and the score on the page. Markup that describes a verdict the page does not show is the quickest route to losing the snippet, and in serious cases to a manual action.

07Why do review stars disappear?

Stars that showed last month and are gone now usually trace back to a short list of causes. Work through them in this order before changing any code.

• The item type is not on Google's supported list, or the page reviews a business it controls.
• The rating or review text is not visible on the page, or the markup describes reviews the page does not show.
• A required property is missing or the scale is wrong, so the rating does not parse as intended.
• The reviews are fake, or were given for a benefit that is not clearly disclosed, which the guidelines forbid outright.
• Google's systems chose not to show the feature for that query, which no markup can override.

The troubleshooting notes in the review snippet documentation are explicit that Google does not guarantee features that consume structured data will show up, and that after a structured data manual action Google ignores the markup on the page, although the page can still appear in results. The Rich Results Test catches syntax problems; it cannot tell you whether the content meets the guidelines.

When the cause is policy rather than syntax, the fix is in the content. Remove reviews you cannot stand behind, show the ones you keep, and make the markup a faithful copy of what is on the page.

08Does review markup matter in AI answers?

Google's guidance on AI features in Search says there is no special schema.org markup needed to appear in AI Overviews or AI Mode, and lists making sure structured data matches the visible text among the basics. Review markup does not buy a place in an AI answer.

What it can do is state the item, the score and the reviewer in a form every parser reads the same way. That is useful to any system trying to work out what a page says, provided the visible review says the same thing.

How AI answers treat ratings, and where they take them from, is a separate question with its own guide: review ratings in AI answers. For review schema itself the conclusion is short. Honest, visible, well-formed reviews are the only kind worth marking up, whoever the reader is.