Can AI Read the Reviews and Creator Content on Your Product Pages?
Adobe scores product pages lowest of any retail page type for AI readability. How to check whether crawlers can see your reviews and creator content.
In April, Adobe scored US retail sites on how much of each page an AI system can actually read. Individual product pages came last of every page type measured, averaging 66%. Homepages averaged 75%. Returns pages scored 82%.
So the page where a shopper decides to buy is the page a machine understands worst. And it is usually where your reviews and creator content live.
Most advice on AI visibility tells you to publish more of that content. We have written about when that is the right call. But if a crawler can't read the reviews and videos already on the page, more of them won't help. Check what can be read before you buy more.
The short answer: reviews and creator content on a product page only help in AI answers if crawlers can read them. That depends on whether the content is in the page's initial HTML, whether it carries structured data, and whether videos have text attached. Audit readability first. Volume doesn't fix a parsing problem.
TL;DR
- 📉 Product pages read worst. Adobe's AI Content Visibility Checker gave US retail product pages an average score of 66%, the lowest page type in its April 2026 benchmark.
- 🛒 The traffic still lands there. Adobe found AI-referred traffic to US retail sites grew 393% year over year in Q1 2026 and converted 42% better than non-AI traffic in March.
- 🏠 Your own site is the cited surface. In LLM Pulse's analysis of 391,073 citations on non-branded shopping questions, about 64.7% pointed to brand and manufacturer sites and 2.9% to all retailers and marketplaces combined.
- 📦 Widgets can hide reviews. Bazaarvoice says review content loaded by client-side scripts may appear as an empty container to some AI crawlers.
- ⏱️ Readable reviews get cited fast, with caveats. Yotpo found review-backed articles had a median of 2 days to first AI citation versus 14 for brand-written posts, and says itself this is not a controlled experiment.
Which product page should you check: yours or Amazon's?
Yours. Readability is a property of how a page is built, and you only control how your own site is built. Amazon, Walmart and Target decide how their listings render, so the lever you hold there is what content exists, not whether a crawler can parse it.
That split matters because the two surfaces do different jobs.
In LLM Pulse's study of non-branded US shopping questions across ChatGPT, Google AI Mode, AI Overviews, Gemini and Perplexity, roughly 64.7% of citations went to brand and manufacturer sites. Every retailer and marketplace domain combined got 2.9%. Retailer listings still convert shoppers who are already there. They just aren't where answer engines usually look.
Which surface you control
- Your own product pagesYou and your developersReadability: initial HTML, structured data, text on videoThe audit below
- Retailer listings (Amazon, Walmart, Target)The retailerPresence: images, brand video, creator video, A+ Content, reviewsCompare against category top sellers
- Syndicated reviews on your siteUsually your review vendor's widgetAsk how the widget loads, and whether it outputs structured dataStep 3 of the audit
When we built Cohley's Amazon Content Benchmarks across 1,777 categories in August 2026, we measured presence: how many gallery images, brand videos and creator videos top sellers carry, and how many use A+ Content. We did not measure readability, because on Amazon it isn't the brand's to change.
This piece is about the other half: the pages you do control.
What can a crawler actually read on a product page?
Mostly what arrives in the first HTML response. Adobe's own guidance says AI agents typically do not execute JavaScript, so content rendered client-side, including product descriptions and user reviews, can be missing from what they see even when a shopper sees all of it.
That is the mechanism behind the 66%. It is also why the parts of a product page brands work hardest on are often the parts crawlers miss.
Here is where reviews and creator content usually go missing:
Where reviews and creator content go missing
- Review widgetReviews injected by a client-side script after page loadAn empty container, per Bazaarvoice
- Star ratingRendered as an image or icon set, with no number in text or markupNo rating
- Creator videoEmbedded player with no transcript, caption text or descriptionA video file it can't interpret
- Photo galleryLazy-loaded carousel, images without alt textFew or no images, no context
- Q&ACollapsed accordions loaded on clickNothing below the question
- Structured dataMissing, or describing content that isn't visible on the pageNo machine-readable product record, or a mismatched one
Sources: Bazaarvoice and Adobe Experience League.
Bazaarvoice's answer to the widget problem is its Authentic Discovery API, which serves reviews, ratings, photos, videos and Q&A in the server-side HTML as Schema.org JSON-LD. Adobe sells a version of the same fix for Adobe Commerce catalogs. Both are fixes to rendering, which tells you where the problem sits.
Creator video has a separate problem. Language models are still mostly text systems, so a video's value to them comes from the words around it: spoken lines captured in a transcript, captions, on-screen text and a specific description. We covered how to brief creators so that text exists in the first place. A video with a three-word caption is close to invisible, however well it converts with people.
What does the research show about reviews and AI citation?
That readable, review-backed content gets cited faster and more widely, in one vendor's data, with confounds the vendor names itself. Nobody has shown that structured data or reviews cause a product to be recommended.
The strongest dataset this fall is Yotpo's. It compared 96 articles generated by its Discover product against 5,686 brand-written blog posts across 15 brands, tracked in ChatGPT, Gemini and Google AI Mode from February 24 to September 22, 2026. Review-backed Discover articles had a median of 33 citations in their first 14 days. Discover articles without reviews had 1.5, and brand-written posts had 0.5. 96% of the review-backed articles were cited at least once, against 40% of brand-written posts.
Read the caveats before you quote it. Yotpo says the result is directional evidence, not a controlled experiment. Discover articles are written around questions AI engines already answer, and brand posts weren't, so topic choice and reviews are tangled together. And it compares Yotpo's own product output against brands' existing content.
It also doesn't test product pages. These are articles. That gap is exactly why the audit below matters: nobody has published how often review and creator content on product pages is readable, so you have to check your own.
What everyone in this market concedes: Bazaarvoice, which sells the structured-data fix, says no brand can guarantee how or when an AI platform will cite or recommend a product. Readability gets your content considered. It doesn't get it chosen.
Stack Influence makes a related point well. Being recommended, being the merchant offer the shopper can buy, and completing an order are three separate outcomes. Fix readability and you have improved the inputs to the first one only.
How do you audit your product pages for AI readability?
Take your ten highest-revenue product pages and check six things. An ecommerce lead and one developer can do it in an afternoon, and none of it needs a paid tool.
- Score each page with a free checker. Adobe's AI Content Visibility Checker is a free browser extension that shows what share of a page AI can read and highlights what people see but machines don't. Record a score per page so you have a baseline.
- Look at the raw HTML. Ask a developer to fetch each page the way a crawler does, before any JavaScript runs. Then search the response for one review sentence, the star rating as a number, and a line from a creator video's caption. If they're missing, crawlers that don't run scripts can't see them.
- Ask your review vendor how its widget loads. Server-rendered or client-side? Does it output Schema.org markup, and for which elements: reviews, ratings, photos, video, Q&A? Get the answer in writing.
- Check the structured data matches the page. Product, Review and AggregateRating markup should describe what a shopper can actually see. Google's guidance, as summarized by Stack Influence, is that structured data should match visible content and that no special AI-only schema is required.
- Give every creator video text. A transcript, a descriptive caption, and on-page text naming the product, variant and use case. If the video shows a specific size or shade, say which in text.
- Check you aren't blocking search crawlers. OpenAI separates OAI-SearchBot, which supports search visibility, from GPTBot, which concerns model training. You can allow one and block the other. Review robots.txt and firewall rules with your developers rather than making blanket changes.
Use this as the scorecard:
AI readability scorecard
- Page readability scoreAdobe AI Content Visibility CheckerProduct pages well below your homepage score
- Review text in initial HTMLRaw HTML fetch, search for a review sentenceReview sentence only appears after scripts run
- Rating as dataRaw HTML and structured dataStars as images, no number anywhere
- Review structured dataInspect JSON-LD for Review and AggregateRatingMissing, or counts that don't match the page
- Creator video textView page text without the playerNo transcript, caption or product description
- Crawler accessrobots.txt and firewall reviewSearch crawlers blocked along with training crawlers
Who this is for: ecommerce and site teams that own a direct-to-consumer storefront.
If you sell only through retailers, skip to presence: whether your listings carry the images, video and reviews top sellers in your category carry.
What should you fix first?
The fix that unlocks content you already paid for. Rendering problems hide every review and video on the page at once, so they come before anything that adds new content.
Fix priority
- 1Get review text and ratings into the initial HTML (server-side rendering, or a vendor option that does it)Medium to highDevelopers and review vendor
- 2Add or correct Product, Review and AggregateRating structured dataLow to mediumDevelopers
- 3Unblock search crawlers you meant to allowLowDevelopers and security
- 4Add transcripts, captions and product text to creator videoLow per videoContent team
- 5Then, if the content is thin, add more reviews and creator contentOngoingEcommerce and content teams
Only the fifth row is a content problem. If you find yourself there first, the earlier post on presence versus representation covers when more content is the right next dollar. Our reviews and UGC programs sit on that side of the line.
Why won't more content fix a parsing problem?
Because a crawler that can't read your first fifty reviews can't read your next fifty either. Volume and readability are separate variables. Adding reviews to a page whose widget renders as an empty container adds nothing a machine can use, however much it helps shoppers.
This is the argument we'd push hardest on with any vendor, including us. If a pitch for more content doesn't start by asking whether your current content can be read, it is selling you the step that comes fifth.
What can nobody promise you?
That readable pages get cited or chosen. Readability makes your content available to answer engines. Whether they use it depends on factors no brand controls and no vendor has published causal evidence for, Cohley included.
Crawler behavior also changes without notice. A page that reads cleanly this quarter can regress after a site redesign, a new widget or a template change. Re-run the checker whenever your product page template changes, and at least quarterly.
Frequently asked questions
- Can AI crawlers read reviews and creator content on my product pages?
Only if the content reaches them. Many AI crawlers don't run JavaScript, so reviews and media loaded by scripts after the page loads can be invisible to them. Adobe's April 2026 benchmark scored US retail product pages at 66% readable on average, the lowest of any page type it measured.
- Does structured data guarantee my product gets recommended by AI?
No. Structured data helps machines read and attribute your product information, but it doesn't decide whether an answer engine recommends you. Bazaarvoice, which sells a structured-data product for reviews, says no brand can guarantee how or when an AI platform will cite or recommend a product.
- Do reviews or creator content matter more for AI visibility?
They solve different gaps. Reviews on your own domain help with representation once you're already present. Off-site creator content helps build presence when you aren't. Either only helps if a machine can read it. Diagnose which gap you have before choosing, and check readability first.
- How do I check whether my product page content is machine-readable?
Score the page with Adobe's free AI Content Visibility Checker, then have a developer fetch the raw HTML before JavaScript runs. Search it for a review sentence, the star rating as a number, and a creator video caption. Anything missing there is likely missing for crawlers too.
- Does syndicated review content on a retailer's site count?
It counts for shoppers on that retailer, but retailers control how their pages render, so readability there isn't yours to fix. In LLM Pulse's study of non-branded shopping questions, retailer and marketplace domains combined drew 2.9% of citations, against about 64.7% for brand and manufacturer sites.
- What's the difference between being cited by AI and being bought through AI?
Being cited means an answer references your product or page. Being bought means a shopper chose your offer and completed an order. Stack Influence treats recommendation, merchant offer and order as three separate outcomes, each with its own failure points. Readability improves inputs to the first one only.
- Should I publish more UGC or fix what I have?
Fix what you have first if the audit shows a rendering problem. A crawler that can't read your existing reviews can't read new ones either. If your pages read cleanly and the content is simply thin, then more reviews and creator content are the right next step.
Next step
Run the six-step audit on your own product pages first. Then do the other half for the listings you don't control: our free Amazon Content Audit compares your images, brand video, creator video and A+ Content against the top sellers in each listing's category. It checks whether the content is there, not whether machines can read it.