Reviews and AI visibility: business ratings, product reviews and schema

Reviews help customers evaluate your business, but Google’s rules differ for business ratings and product reviews. Learn what to mark up, how to collect feedback fairly, and why good reviews still need clear prices and a working booking or checkout path.

Reviewed October 1, 2026 · 7 min read

Reviews can help people and AI-assisted search systems evaluate a business. To use them well, distinguish three things: testimonials on your site, reviews on a platform such as Google Maps, and structured data intended for search features. The rules for one do not automatically apply to the others.

Google’s review snippet guidelines exclude self-serving reviews of LocalBusiness and Organization entities from the star-review feature. That includes a business displaying reviews about itself through a third-party widget. This is a rich-result eligibility rule, not a ban on publishing genuine testimonials.

Product review markup and Merchant Center’s Product Ratings program have their own requirements. A shop can collect customer reviews of products it sells, but that does not let it present its own business testimonials as product ratings. Identify what is being reviewed before choosing the markup or submission route.

Check what your review markup describes#

Use this review before adding or changing rating schema:

• Identify the subject. Is the review about your business, a specific product or another supported item? Do not change the type simply to obtain stars.
• Check the applicable eligibility rules. Reviews of your own LocalBusiness or Organization do not qualify for Google’s star feature on your site.
• Keep marked-up reviews visible and accurate. The rating, count and reviewed item should match what visitors can see.
• Validate the implementation. Use the relevant feature documentation and Rich Results Test, then inspect Search Console for detected issues.

For eligible review markup, required fields depend on whether the review is standalone or nested within the reviewed item. Use the current documentation rather than a generic plugin checklist. The JSON-LD guide explains how structured data should reflect visible page content.

Where assistants may find customer feedback#

Sources vary by assistant and query. Google explains that its local listings combine several sources, including public websites, licensed data and user contributions. Keep your own records clear, while recognizing that independent reviews are outside your editorial control.

Do not rewrite customer reviews to make them sound more favorable, invent feedback or hide dissatisfied customers behind a different submission path. A review process should invite honest experiences and preserve the distinction between customer comments and your own claims.

Google can summarize review themes, so a customer may see recurring comments about communication, waiting times or value alongside a rating. Other assistants may quote or summarize different sources. The local AI visibility guide explains why consistent business facts matter across those sources.

Why recent reviews matter#

Google says its local review summaries use reviews from the past year and are updated regularly. That describes the summary feature, not a rule that older reviews stop counting everywhere. Recent feedback can help customers understand how the business operates now.

Build a consistent, voluntary request for honest feedback into your normal follow-up. Invite customers fairly, without choosing only people you expect to leave a positive rating. Read the responses for problems you can fix, such as unclear arrival windows or confusing cancellation terms.

Product Ratings has a separate submission process#

Google’s Product Ratings program can show aggregated product feedback in shopping ads and free listings. Its requirements include:

1. At least 50 reviews across your products to participate. This is not a universal minimum for every type of review markup.
2. A supported submission route, through a review aggregator or direct enrollment and submission in Merchant Center.
3. Accurate product matching. Use valid identifiers such as GTINs where available, with consistent brand and MPN data as applicable.
4. Complete, accurate review-feed updates at least monthly, following the program’s content policies.
5. Ongoing diagnostics. Meeting the minimum does not guarantee that ratings will display on every product or search.

The Product Ratings policies require comprehensive updates and prohibit removing old reviews merely because they are old. They also specify how permitted incentivized product reviews must be disclosed. Keep product and review identifiers aligned with the catalog described in the product-feed guide.

Collect reviews without distorting them#

1. Audit the subject and source of each rating on your site. Correct anything that combines unrelated products, businesses or review counts.
2. Keep your business profile accurate so customers review the right location and service.
3. For Google Maps reviews, follow its review policy: do not offer incentives or selectively solicit positive feedback. A disclosure does not make an incentivized Maps review acceptable.
4. For product reviews, follow the destination’s separate policy. Where an incentive is permitted, it must not depend on a positive opinion and must be disclosed as required.
5. Respond to feedback professionally and investigate recurring issues. Preserve customer privacy when discussing a problem publicly.

Connect a good reputation to a useful next step#

Reviews can support a customer’s decision, but they cannot answer every practical question. A well-reviewed service still needs to explain where it operates, what it offers, how pricing works and how to check availability. A product still needs accurate stock and delivery information.

Check the journey after the recommendation. Can someone request a quote, book a suitable time or buy the selected item? AI calls to local businesses illustrate what happens when an assistant has to seek information that is missing online.