Review Schema
The review schema is a structured data block in the schema.org standard with which you mark up customer reviews and star ratings in a machine-readable way. It tells search engines and AI systems who reviewed what, how many stars were given and how many people voted. This turns reviews into reliable, readable data instead of mere prose.
Why it matters
Reviews are one of the strongest trust signals of all - but only if machines can capture them cleanly. If your 4.7-star rating appears only as prose on the page, a system has to guess whether it is a real rating. With the review schema you deliver the number as an unambiguous data field. In classic search results this often produces golden stars (a so-called rich snippet, i.e. an enriched search result). This noticeably increases the click-through rate. For AI systems like ChatGPT or Perplexity, the rating becomes a fact they can cite when users ask for the best provider. Without markup, your good reputation stays invisible to machines.
How it works
Technically, you usually write the review schema as JSON-LD, a small script in the page's source code that stays invisible to visitors. In it you define two building blocks: a single review (Review) with author, text and star value, or an aggregated overall rating (AggregateRating) with an average score and the number of reviews. Every rating needs a reference: it belongs to a product, a company, a recipe or an event. The rating scale is important - Google expects 1 to 5 by default. Whoever uses a different scale must state the highest and lowest value. You can check the whole thing with Google's Rich Results Test, which shows errors directly.
Common mistakes
The gravest violation is invented or self-assigned ratings. Google expressly forbids you to review your own company or invent ratings that are not even visible on the page. The schema must always reflect real, visitor-readable reviews. Another classic: stars without a number of reviews look unbelievable and are sometimes ignored. Equally problematic is marking up an overall rating without a concrete reference object. Outdated markup also does harm - if the page shows five reviews but the schema claims five hundred, manual penalties loom. Always keep the data block and visible content in sync, otherwise you lose trust with search engines and AI alike.
Relevance to AI recommendations
When someone asks an AI assistant "Which tax advisor in Cologne is worth recommending?", the system looks for solid signals. A cleanly marked-up overall rating with many votes is such a signal: it is unambiguous, quantifiable and hard to overlook. AI models increasingly draw on structured data because it leaves less room for interpretation than prose. This makes the review schema a building block of your AI visibility: it increases the likelihood of being named and classified positively in a generated answer. Combined with honest, current reviews, a trust anchor arises that both search engines and generative systems like to pick up and pass on.
Example
A small coffee roastery sells its beans online. Each variety has customer reviews. Until now the stars appeared only as a graphic on the page. The operator now adds a review schema as JSON-LD for the variety "Espresso Forte": average 4.6 out of 5, based on 83 reviews. A few days later Google shows golden stars directly below the product link in the search result. The click-through rate rises. If someone later asks an AI assistant for a strong espresso for the portafilter machine, the roastery appears with its strong rating as a concrete recommendation.
Common questions
May I mark up my own company with the review schema?
No. Google prohibits self-reviews as well as invented ratings. The schema may only depict real reviews from customers or third parties that are visible to visitors, otherwise manual penalties loom.
What is the difference between Review and AggregateRating?
Review is a single review with an author and text. AggregateRating is the summarized overall score from several reviews, i.e. the average value plus the number of votes cast.