Aizii Learn · AI Search

[← Back to Learn](https://aizii.ai/learn)

# Do third-party reviews affect AI recommendations

Yes, third-party reviews affect AI recommendations, because they are the words an assistant has to quote when it explains why it picked one product over another. Your own site can state your price, your materials, and your warranty. It cannot state, in a way anyone will trust, that your product is comfortable, durable, or worth the money. That claim has to come from someone else, in public, where an assistant can read it and check it. A brand with no third-party reviews gives the assistant nothing to cite but its own marketing copy, and assistants prefer a stranger's experience to a seller's promise.

The effect is indirect, and that matters. No vendor publishes a formula that turns a star rating into a recommendation, and no review count buys a place in an answer. Reviews work by supplying the evidence an assistant needs: specific experiences, named pros and cons, and a pattern across many buyers that lets the assistant describe the product in the words buyers actually use. A hundred vague reviews are worth less than ten detailed ones. What follows is how to earn the second kind, and how to keep the first kind from hurting you.

## Where it goes wrong

Reviews go wrong in a few specific ways, and most of them come from treating reviews as marketing instead of evidence.

-   The brand writes its own reviews, or pays for them. The FTC's consumer reviews rule bans fake or false reviews, including reviews from people who never used the product, reviews bought from others, and reviews procured from company insiders without disclosure. It also bans paying for a review on the condition that it expresses a particular sentiment, positive or negative. A bought five-star review is not a review. It is a violation, and it is also useless as evidence, because an assistant that quotes it is quoting an ad.
-   The brand hides the bad reviews. The same rule bans suppressing reviews, including hiding negative ones while showing positive ones. A review page that shows only praise tells an assistant, and a shopper, that the page cannot be trusted. Keep the negative reviews visible, answer them, and let the pattern speak.
-   The brand marks up reviews that are not on the page. Google's review snippet guidelines require that the review content you mark up is readily available to users from the marked-up page. If your structured data claims a rating and the page shows no reviews, the markup is unsupported, and Google may ignore it. The same rule applies to aggregate ratings: if you mark up an aggregate rating, users should be able to see that aggregate rating on the page.
-   The brand aggregates reviews from other sites into its own markup. Google's guidelines are direct on this point: do not aggregate reviews or ratings from other websites. Reviews collected on a marketplace or a review platform belong to that platform. Copying them into your own markup as if they were yours is not allowed, and it will cost you the snippet.
-   The brand reviews itself. Google's guidelines bar a local business or an organization from earning review snippets for reviews about itself. Reviews of your own business, hosted on your own site, cannot earn you stars in search, even through a third-party widget. Product reviews are different: a product page can carry reviews of the product, and that is where the markup belongs.
-   The brand asks for reviews only from happy customers. Asking only the buyers you know loved the product skews the pattern. Ask every buyer, in the same words, after the same experience, and let the rating be whatever it is. An incentive, if you offer one, must never depend on the review being positive, and it must be disclosed clearly, because a rewarded review that hides the reward is deceptive.
-   The brand never answers a review. An unanswered complaint is a dead end for a shopper and for an assistant. An answered complaint, with a fix or an explanation, is evidence that the brand stands behind the product. The answer is part of the review now, and assistants quote answers too.

## The steps

**Step 1. Ask every buyer, in the same words.** The marketer sets up a post-purchase request that goes to every buyer after delivery, with the same wording and the same timing, and no filter for who had a good experience. The request asks for a review of the product, not of the brand, and it asks for specifics: comfort, fit, durability, and what the buyer would tell a friend. A review flow that only reaches happy buyers produces a rating nobody trusts.

**Step 2. Collect reviews where buyers already look.** The marketer lists the review surfaces that matter for the brand: the product pages on the brand's own site, the marketplaces where the product is sold, and the independent review platforms buyers check before they buy. Do not try to pull every review onto your own site. Let each surface keep its own reviews, and link to the surfaces from the product page so shoppers, and assistants, can follow the trail.

**Step 3. Put the reviews on the product page, where the markup can see them.** The developer makes sure the product page shows the reviews it marks up: the review text, the author, and the rating, visible to any visitor without a login. Google's guidelines require the marked-up content to be visible on the page, and an assistant can only quote what it can read. A review hidden behind a tab that loads on click, or behind a login, is a review that does not exist to the reader.

**Step 4. Mark up the aggregate rating, not a single cherry-picked review.** The developer adds AggregateRating structured data to the product page, with the average rating and the count of reviews, matching the numbers shown on the page. Google's guidelines say that for aggregate reviews you must supply the average rating for the rich snippet to display, and that if you mark up multiple individual reviews you should include an aggregate rating alongside them. Do not mark up a rating for a category page, a list page, or a page about the brand itself. One product per page.

**Step 5. Disclose every incentive, and never tie one to sentiment.** The marketer checks every review flow for rewards. If a buyer gets a discount, a free product, or loyalty points for leaving a review, the review must say so clearly, and the reward must be given whatever the review says. A reward that depends on a positive review is banned by the FTC's rule, even if the review discloses it. An undisclosed reward is deceptive under the FTC's endorsement guides, which require a material connection between the reviewer and the brand to be disclosed clearly. When in doubt, drop the incentive and ask for the review plain.

**Step 6. Answer the negative reviews first.** The marketer answers every negative review on the brand's own pages with a specific reply: what went wrong, what the brand will do, and how the buyer can reach a person. Do not argue, do not blame the buyer, and do not copy the same reply to every complaint. A specific answer turns a complaint into evidence of service, and it gives an assistant something fair to quote alongside the complaint itself.

**Step 7. Watch what the assistants say, and fix the source, not the answer.** The marketer checks the brand's core questions in the assistants each month and notes which reviews get quoted and which facts get stated wrong. When an assistant quotes an old review about a discontinued version, the fix is on the product page and in the review flow, not in a complaint to the assistant. Update the page, answer the old review with the current facts, and let the new pattern replace the old one.

## The machine-readable piece

This is the Product structured data for a product page with an aggregate rating, following Google's review snippet guidelines. It states the product, the average rating, and the number of reviews, and nothing else. The numbers must match what the page shows. Replace the placeholders with your own product details, replacing yourdomain.com with your actual domain.

```
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Summit Latex Pillow",
  "description": "A latex pillow with a removable cotton cover, in standard and king sizes, with a choice of low and high loft.",
  "url": "https://[yourdomain.com]/products/summit-latex-pillow",
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": "4.6",
    "bestRating": "5",
    "reviewCount": "214"
  }
}
</script>
```

Check the block with the rich results test Google documents, and confirm the page shows the same rating and the same count, with the review text visible to any visitor. Do not add a rating value the page does not show, do not count reviews from another website, and do not mark up a rating for the brand itself. A rating that matches the page is evidence. A rating that does not match is a reason to ignore the markup.

## What it looks like when it works

When reviews work, the assistant's answer sounds like a buyer, not like a brochure. It names the product, gives the price from your page, and then quotes what buyers said: that the pillow sleeps cool, that the high loft suits side sleepers, that the cover washes well. The recommendation carries weight because the reasons come from people who bought the product, and the shopper can follow the link to the reviews and read them. The brand did not write the recommendation. It earned the evidence the recommendation is built from.

Say you run a bedding brand, selling pillows direct from your site. Your product pages state the materials and the sizes, but your brand rarely appears in AI answers for sleep questions, and when it does, the assistant quotes a competitor with hundreds of reviews. The marketer sets up a post-purchase request that goes to every buyer, with no filter for happy customers. The developer puts the reviews on the product page, visible without a login, and adds the aggregate rating markup with numbers that match the page. The marketer answers the negative reviews, including one about a pillow that arrived too firm, with a specific reply and an exchange. Months later, the pattern is real: most buyers praise the cooling cover, some note the firmness, and the average reflects both. An assistant answering a question about pillows for hot sleepers names your pillow, quotes the cooling cover in buyers' words, and notes the firmness as the tradeoff. The reviews did not buy the recommendation. They gave the assistant the words to make it, and the words were true.

### The fundamentals

These mechanics are shared across every engine and covered in full in their own guides:

-   •The llms.txt format and a complete sample, see our [guide to writing a good llms.txt file](https://aizii.ai/learn/ai-search/how-to-write-a-good-llms-txt-file)
-   •The markup AI systems read to verify facts, see our [product schema guide](https://aizii.ai/learn/ai-search/how-to-add-product-schema-markup-for-ai-search)
-   •The complete crawler rule reference, see our [robots.txt guide for AI search](https://aizii.ai/learn/ai-search/how-to-write-robots-txt-for-ai-search)

## What this guide does not cover

This guide covers how third-party reviews feed AI recommendations and how to earn them within the rules. It does not cover the full technical audit, or the transaction layer that lets an agent complete a purchase, and it does not cover review platforms' own policies beyond the vendor rules cited here, so check each platform's rules before collecting reviews there. It is not legal advice: the FTC rules cited here are summarized in plain words, and a lawyer should review your review flows before you scale them.

## Check your own store now

Run the free scan. One click, no signup, and it shows whether AI crawlers can read your store today: aizii.ai/aeo-scan

[Run the free scan](https://aizii.ai/aeo-audit)

## Sources

-   Google Search Central, Review snippet structured data (specific): [https://developers.google.com/search/docs/appearance/structured-data/review-snippet](https://developers.google.com/search/docs/appearance/structured-data/review-snippet)
-   Google Search Central, Product snippet structured data (specific): [https://developers.google.com/search/docs/appearance/structured-data/product-snippet](https://developers.google.com/search/docs/appearance/structured-data/product-snippet)
-   Schema.org, Review type (specific): [https://schema.org/Review](https://schema.org/Review)
-   Schema.org, AggregateRating type (specific): [https://schema.org/AggregateRating](https://schema.org/AggregateRating)
-   FTC, Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465 (specific): [https://Business.FTC.gov/system/files/ftc\_gov/pdf/2025-Fake-Review-Warning-Template.pdf](https://business.ftc.gov/system/files/ftc_gov/pdf/2025-Fake-Review-Warning-Template.pdf)
-   FTC, Endorsement Guides, 16 CFR Part 255 (specific): [https://www.govinfo.gov/content/pkg/CFR-2026-title16-vol1/pdf/CFR-2026-title16-vol1-part255.pdf](https://www.govinfo.gov/content/pkg/CFR-2026-title16-vol1/pdf/CFR-2026-title16-vol1-part255.pdf)

Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.
