AI-written reviews and review snippets

01What counts as a fake or AI-written review?

AI-written reviews are reviews whose text was produced by a generative model rather than typed by the person named as the reviewer. Some are invented from nothing: no customer, no purchase, no experience. Others start from a real customer whose rough notes were tidied by a writing tool. The two cases look the same on the page, and they are not the same thing.

A fake review is a review that misrepresents who wrote it or what they experienced. That definition matters more than the tool used. The FTC's final rule, announced in August 2024, describes it in exactly those terms: it targets reviews that misrepresent that they are by someone who does not exist, such as AI-generated fake reviews, or by someone who did not have actual experience with the business or its products, or that misrepresent the experience of the person giving it.

Google's review snippet guidelines draw the same line from the search side. A review that is not based on a genuine experience of a product or service is treated as fake, whoever or whatever wrote the words. Neither source, as read in November 2026, says a genuine customer who polishes their own wording with a tool has written a fake review. Both focus on identity and experience.

02What do Google's review snippet policies say?

A review snippet is the star rating and short review summary that Google can show next to a search result when a page carries valid Review or AggregateRating markup. Google's review snippet documentation sets content rules on top of the technical ones, and several of them are about authenticity rather than syntax.

• The reviews you mark up must be readily visible to users on the marked-up page, with the review text and rating on show.
• Reviews and ratings must not be aggregated from other websites.
• Fake or undisclosed incentivized reviews must not appear on the page or in the structured data.
• The reviewer's name must be a valid name, and a promotional phrase is not one.
• If the entity being reviewed controls the reviews about itself, its pages that use LocalBusiness or another Organization type are ineligible for star review features, including reviews pulled in through an embedded third-party widget.

The self-serving rule is older than generative AI. Google announced it in a September 2019 post on making review rich results more helpful, which also narrowed the list of schema types that can trigger review stars. That change is why a business marking up praise for itself on its own services page sees no stars, however the reviews were written.

Google's general structured data guidelines add two rules that matter for review pages. Markup must not describe irrelevant or misleading content, and the guidelines name fake reviews as an example. If a page shows several reviews, the markup should include all of the reviews that are visible, because marking up only the best ones would mislead people who expect the page to match its search appearance. A site that breaks these rules risks a manual action; the same guidelines explain that a structured data manual action removes eligibility for rich results but does not affect how the page ranks in web search.

03What does the FTC rule on fake reviews (2024) prohibit?

The US Federal Trade Commission announced its final rule banning fake reviews and testimonials on 14 August 2024. The Commission voted 5 to 0, and the announcement says the rule lets the agency seek civil penalties against knowing violators. The announcement says the rule becomes effective 60 days after its publication in the Federal Register. The rule is US law, so check local rules if you sell elsewhere, and treat this summary as orientation rather than legal advice.

The announcement lists six groups of prohibited practices. In plain terms:

• Fake or false reviews and testimonials: creating or selling them, and buying or disseminating them when the business knew or should have known they were fake.
• Buying positive or negative reviews: offering compensation or incentives conditioned on a review expressing a particular sentiment, whether the condition is stated or implied.
• Insider reviews: reviews by officers, managers, employees or agents that do not clearly disclose the material connection, plus rules on soliciting reviews from relatives and staff.
• Company-controlled review sites: presenting a site you control as an independent reviewer of a category that includes your own products.
• Review suppression: using groundless legal threats, intimidation or false accusations to remove negative reviews, or implying the reviews you show are all or most of those submitted when some were suppressed for being negative.
• Fake social media indicators: buying or selling bot-generated followers or views to misrepresent influence for a commercial purpose.

Two of these overlap directly with markup. The suppression rule mirrors Google's requirement to mark up all visible reviews, and the ban on fake reviews mirrors Google's ban on fake reviews in structured data. A review program that satisfies one is most of the way to satisfying the other.

04How do you screen AI written reviews before you publish them?

Screening is the step between a review arriving and a review being shown and marked up. Both sources turn on whether a real person had a real experience, so screening works best when it checks for evidence of that experience rather than trying to judge prose style. The steps below are a workable baseline for a WordPress or WooCommerce site.

1. Tie each review to a real transaction, booking or account where your platform allows it, and record that link privately.
2. Ask for a name the reviewer is willing to show and a rating with a written comment, which Google's guidelines recommend but do not require.
3. Hold reviews for moderation instead of publishing them on submission, so nothing reaches the page or the markup unchecked.
4. Reject reviews from staff, agents or relatives unless the connection is disclosed in the review itself.
5. If you offer an incentive for reviewing, offer it for any honest review and disclose it clearly and prominently next to the review.
6. Never delete or hide a genuine review because it is negative; remove only reviews that break a published content policy, and keep a record of why.
7. Publish your review policy on the site so customers and moderators work from the same rules.

Customers increasingly attach photos to reviews, and some of those images will be generated or edited. Images carry their own labels, which are covered in IPTC digital source type labels. If your own team uses AI to summarize reviews or draft replies, say so on the page; disclosing AI-assisted content has a pattern for that.

05Markup only what is real

Review markup is a claim about the reviews on the page. Once screening is done, keep the JSON-LD a faithful copy of what is visible: the same reviewer names, the same ratings, the same dates, and an aggregate that is computed from every review shown rather than the best few. For the full property list, see how review snippets work in the Schema types section.

1{2  "@context": "https://schema.org",3  "@type": "Product",4  "name": "Example Kettle",5  "aggregateRating": {16    "@type": "AggregateRating",7    "ratingValue": "4.5",8    "reviewCount": "2"9  },10  "review": [11    {12      "@type": "Review",13      "author": {14        "@type": "Person",15        "name": "Example Customer"216      },17      "datePublished": "2026-10-02",18      "reviewRating": {19        "@type": "Rating",20        "ratingValue": "5",21        "bestRating": "5"22      },23      "reviewBody": "Boils quickly and the lid seals well. Bought it in September."24    },25    {26      "@type": "Review",27      "author": {28        "@type": "Person",29        "name": "Second Example Customer"30      },31      "datePublished": "2026-10-19",32      "reviewRating": {33        "@type": "Rating",34        "ratingValue": "4",35        "bestRating": "5"36      },37      "reviewBody": "Works well, a little loud. Received a discount code for reviewing."338    }39  ]40}

Leave out anything you cannot stand behind. If a review fails screening, remove it from the page and the markup together. If a review widget pulls ratings from another website, do not copy those ratings into your own Review markup, because Google's review snippet rules forbid aggregating reviews from other websites. And if you are a local business showing reviews of yourself, expect no stars from that markup even when every review is genuine.

Honest review data also matters beyond search results. AI answer engines read the same pages, and a page whose markup matches its visible reviews gives them nothing contradictory to work with. For how review data surfaces in answers, and how the signals in this section connect to the rest of this handbook, start from the rest of this handbook or the content authenticity signals section page.