On July 24, 2026, Google added a guideline to its review snippet documentation saying you must not include fake or undisclosed incentivized reviews on your page or in your structured data markup.
It applies to any site that marks up customer reviews with Review or AggregateRating, including product, service, organization, local business, recipe and software pages. The rule covers the review text visitors can read and the JSON-LD together, so a review that breaks it is a problem in both places.
The first step is an audit, not a code change. List every review on the page, note where each one came from, remove the entries that break the rule, then update the rating and the counts in your markup so they match the reviews that remain.
What Google added on July 24, 2026
Google's documentation updates log records the July 24 entry as "Added a new review snippet guideline." It describes the change as a new guideline in the review snippet documentation "about fake and undisclosed incentivized reviews."
The guideline itself, in Google's review snippet documentation, reads: "Don't include fake or undisclosed incentivized reviews on your page or in your structured data markup." It sits in the documentation for review snippets in Google Search, so it sets what review content and its markup must be to qualify for that feature.
The change is about eligibility and quality for Google Search. It does not make the schema.org review vocabulary invalid, and Review and AggregateRating remain valid types. It also does not mean every existing review display disappeared on July 24. It is a guideline update, and Google separately says it does not guarantee that eligible structured data will appear in Search.
The two cases Google names
Google gives two examples of what the rule covers. The first is a review that is not based on a genuine experience of the product or service. The second is a review written in exchange for money, discounts, vouchers or free products without a clear and prominent disclosure of the incentive.
Read the second case closely. The word that matters is "undisclosed." A review written after a customer received a free product, with the incentive clearly and prominently disclosed beside it, is the situation Google's wording leaves room for. Disclosure addresses only that one issue, though. Every other Google guideline, the rules of the platform you collect reviews on, and the law that applies to you still apply.
Google's July 24 wording does not use the words "filtered" or "gated." Still, Google's general structured data guidelines say markup must be a true representation of the visible page and must not be misleading. If your page shows only a hand-picked set of reviews, say plainly that it is a selection, and do not mark it up as though it were the whole customer population.
One more condition sits beside both cases: review content in your markup must be visible and readily available to visitors on the page. A genuine review that appears only in JSON-LD does not meet that requirement.
Where the older self-serving review rule fits
Google announced a separate review restriction on September 16, 2019, in its post Making Review Rich Results more helpful. It says reviews that a local business or organization controls about itself, including through a third-party widget, are considered self-serving and generally cannot earn the star review feature. That holds even when the reviews are genuine, and it applies to both Review and AggregateRating.
So, there are two different questions to ask of a page. The July 24 rule asks whether each review is genuine and, if incentivized, disclosed. The 2019 rule asks who controls the reviews and what the markup is about. A page can pass the first and still not be eligible for stars under the second.
For a local business, that means your genuine reviews can stay on the page for visitors to read, which is useful in itself. Just do not expect the markup around them to produce stars, and do not rewrite or aggregate reviews to try to get around the restriction. The Schema Nova guide to types of schema markup shows which types describe a business and which describe the things it sells.
Audit the visible reviews and the markup
Think of a small salon owner who has collected reviews for years from walk-ins, a booking app, a social page and a few emails she pasted onto the site by hand. She will not know which ones need attention until she lists them. The audit below works the same way for a shop, a software page or a recipe site.

Work through these steps in order, one page at a time:
- List every review and rating visible on the page.
- Note where each review came from: a direct submission, a verified customer platform, an embedded widget, an imported feed or content entered by hand.
- Remove reviews that are fabricated, written by people with no genuine experience, copied from another product or service, or written in exchange for a benefit that is not disclosed.
- Ask whether any incentive was conditioned on a positive or negative rating, and if an otherwise genuine incentivized review stays, add a clear, prominent disclosure where the platform and the law allow the practice.
- Remove or correct markup for any review that is no longer visible on the page.
- Recalculate
ratingValue,reviewCountandratingCountfrom the reviews that remain. - Confirm that the visible text, author, rating and date of each review match the JSON-LD, and that the aggregate values match the visible set.
- Leave out ratings copied from other websites. Do not fold them into an aggregate rating your own site controls.
Removing a review is a content decision, and the markup follows it. Do the content first, then regenerate or edit the JSON-LD so it describes the page as it now reads.
Before and after JSON-LD
The examples below are teaching code on the reserved example.com domain, with fictional names. Compare them against your own markup, and do not paste them as they are. Replace every name, URL, date and rating with the values from your own visible reviews.
The first example is syntactically valid JSON-LD that still breaks Google's content guideline, because the single review in it is fabricated and the aggregate rating counts it. Valid syntax and acceptable content are separate questions.
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Example Desk Lamp",
"url": "https://example.com/products/example-desk-lamp",
"review": {
"@type": "Review",
"author": {
"@type": "Person",
"name": "Jordan Lee"
},
"datePublished": "2026-07-20",
"reviewBody": "This lamp transformed my office.",
"reviewRating": {
"@type": "Rating",
"ratingValue": 5,
"bestRating": 5,
"worstRating": 1
}
},
"aggregateRating": {
"@type": "AggregateRating",
"ratingValue": 5,
"reviewCount": 1,
"bestRating": 5,
"worstRating": 1
}
}A second example would look almost identical. Imagine a review by a fictional Casey Morgan, dated 2026-07-21, with the text "Bright and easy to install," that was written in exchange for a discount nobody disclosed. The markup is just as valid, and it is just as unsuitable to keep on the page or in the JSON-LD.
In both cases, take out the review and recalculate the aggregate from whatever genuine, visible reviews remain. If none remain, remove the review and aggregateRating properties instead of leaving unsupported values behind. The version below shows a page whose one remaining review is genuine and visible.
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Example Desk Lamp",
"url": "https://example.com/products/example-desk-lamp",
"review": {
"@type": "Review",
"author": {
"@type": "Person",
"name": "Taylor Rivera"
},
"datePublished": "2026-07-18",
"reviewBody": "The adjustable arm makes it easy to aim light at my notebook.",
"reviewRating": {
"@type": "Rating",
"ratingValue": 4,
"bestRating": 5,
"worstRating": 1
}
},
"aggregateRating": {
"@type": "AggregateRating",
"ratingValue": 4,
"reviewCount": 1,
"bestRating": 5,
"worstRating": 1
}
}
Here the visible review and the markup say the same thing, and the aggregate rating describes only the review a visitor can read. Product has a Google review snippet feature, but eligibility is never guaranteed. For the markup to be eligible, Google lists these properties as required for this Google feature: the review author, the reviewRating with its ratingValue, and the name of the reviewed item. For an aggregate rating, it lists ratingValue plus at least one of ratingCount or reviewCount. The datePublished, bestRating and worstRating values are recommended.
Where a schema.org type has no matching Google feature, the markup can still describe the page for other readers, but do not read Google's feature requirements into it. The guide to JSON-LD schema examples has more complete patterns for writing markup by hand.
What Google says can happen to markup that breaks the rule
Google states that violating its review snippet guidelines may lead to a manual action. A manual action is a penalty or restriction Google applies when it determines that a page or site violates its policies. Google's documentation says that when a structured data manual action applies, the structured data on that page is ignored, although the page may still appear in Google Search.
Violations can also keep otherwise valid markup from appearing as a rich result. These are Google's statements about what may happen, and not outcomes you should expect or plan around. If your markup describes only genuine, visible reviews, there is nothing for the rule to catch.
Keep review markup accurate for Bing and AI search
Bing's structured data documentation says it supports Schema.org markup including JSON-LD and processes review and aggregated-review annotations, so the same block serves both Google and Bing. Its Webmaster Guidelines say misleading structured data may be ignored and can affect trust. Yandex's page on working with reviews says it monitors the credibility of review data.
To let ChatGPT search and Perplexity reach the page, allow OAI-SearchBot (ChatGPT search help) and PerplexityBot (crawler announcement) in robots.txt. Across all of them, the audit above is the work: publish only genuine, visible reviews and keep the markup accurate to the page. For the full comparison, see whether AI systems use schema markup.
A short legal backdrop
The U.S. Federal Trade Commission's Trade Regulation Rule on the Use of Consumer Reviews and Testimonials took effect on October 21, 2024. The FTC's announcement of the final rule describes prohibitions on fake or false reviews, on compensation conditioned on a particular positive or negative sentiment, and on certain other practices.
The FTC separates an incentive for an honest review from an incentive tied to a particular sentiment. Google's structured data rule is not a substitute for legal compliance, and the FTC's questions and answers page is the place to read how the rule applies to a specific situation.
Test the corrected page
After you remove the entries and update the counts, test the page in Google's Rich Results Test. A pass there shows that the markup parses and meets the feature's property requirements. It does not show that the reviews are genuine, and it does not predict whether Google will show anything.
After Google recrawls the page, inspect it in Search Console. You can also run the JSON-LD through the free Schema Nova validator to catch syntax problems before you publish. If you need to rebuild the markup from scratch, the review schema generator writes Review and AggregateRating JSON-LD from the details you enter, and you remain responsible for entering only reviews that appear on your page.
Questions people ask
What did Google change on July 24, 2026?
Google's documentation updates log records a new guideline in the review snippet documentation. It says not to include fake or undisclosed incentivized reviews on your page or in your structured data markup.
Can I mark up a review that was written in exchange for a discount?
Google's rule is about undisclosed incentives. If the incentive is not clearly and prominently disclosed, leave the review off the page and out of the markup; if you keep it, disclose the incentive where the platform and the law allow it. Disclosure does not make a review automatically eligible for Google stars.
Does the rule cover visible reviews as well as JSON-LD?
Yes. Google says the guideline applies to your page and to your structured data markup. Marking up a review that is hidden or no longer on the page is not a way around it.
Does removing a fake review remove my whole page from Google?
Google's documentation says that under a structured data manual action the markup is ignored, although the page may still appear in Search. That describes Google's own enforcement. Removing a review changes what your page and markup say, and it makes no promise about what Google displays.
Can I aggregate reviews copied from Google, Facebook or another site?
Do not fold ratings copied from other websites into an aggregate rating that your own site controls. Build the aggregate only from reviews visitors can read on that page. For a local business or organization, the separate self-serving rule from 2019 may still keep star features from appearing.
Is Schema.org review markup still valid if Google does not show stars?
Yes. Schema.org vocabulary and Google feature eligibility are separate things. Review and AggregateRating stay valid types, and other consumers such as Bing document support for review annotations.
