Why AI Won't Recommend Your Product Page: Price, Availability and the Facts Buyers Need

According to Google, its Shopping Graph holds over 50 billion product listings (Pichai, 2026), and your product can get in with no feed and no schema at all.

Google's Merchant Center help says crawled products can show up "even if the information hasn't been marked up using schema.org or added in Merchant Center" (Merchant Center Help, n.d.). So your product page is the base layer, and a feed adds coverage and control. Our GEO for ecommerce guide and AI SEO hub cover the rest of the playbook.

Where do ChatGPT, Google and Copilot get product facts?

From feeds or catalogues that merchants send, and from the open web, including your product pages. Google's Shopping Graph draws on Merchant Center and on "what retailers and brands post across the web" (Rockinson, 2023). Microsoft says Copilot uses "both information found on the web and from a merchant's feed in the Microsoft Merchant Center" (Microsoft, n.d.).

Engine	Where product facts come from	Feed required?	In-chat checkout
ChatGPT	First- and third-party data, plus crawling	No	None: buyers finish on your site. Shopping is US-only
Google AI Mode and Gemini	Merchant Center feeds plus the web	No	UCP checkout: US, Canada, Australia
Microsoft Copilot	The web plus Microsoft Merchant Center feeds	Not stated	Copilot Checkout: US buyers, English, USD

Do you need a product feed for ChatGPT Shopping?

No. OpenAI's merchant FAQ says "No, but feeds give you greater control over how your products appear", and Shopify and Etsy catalogues are already connected (OpenAI, n.d.). Direct feeds are open to approved partners only (OpenAI, n.d.).

In Profound's tracking, the share of ChatGPT Shopping picks classed as feed-sourced jumped from 8.26% to 61.54% on 10 July 2026 (Profound, 2026). It reached about 65% by 3 September (Southern, 2026). That's one vendor's data: July's sample was 1,757,723 prompt runs tracked for Profound's own customers, and Profound sells that tracking.

Which product facts do the engines say they use?

OpenAI says ChatGPT ranks merchants on "availability, price, quality, and whether they are the maker or primary seller" (OpenAI, 2026). The prices it shows come "from third-party providers".

Google's merchant listings cover pages "where customers can purchase products from you" (Google, 2025). They need a name, an image and an Offer with a price above zero in a three-letter ISO 4217 currency (Google, 2026). Google calls availability recommended, but OpenAI's feed spec makes it one of nine required fields, alongside price and brand (OpenAI, n.d.).

Can a UK shop use any of this?

You can be found, but UK buyers can't check out inside any of the three. Shopify's ChatGPT channel takes stores based outside the US if they sell to US customers (Shopify, n.d.). OpenAI is also "moving away from a standalone Instant Checkout experience" (OpenAI, n.d.).

So if you're a UK shop, today's work is on the product page.

Why does the price need to be in your page's HTML?

Because price sways what AI picks in two controlled tests, and some AI systems miss a price that JavaScript adds in the browser. For merchants, Google recommends "putting Product structured data in the initial HTML for best results" (Google, 2026).

Does price change what AI picks?

In two controlled tests, yes. Rahul Vishwakarma and two Sprinklr colleagues ran 252,000 trials in which six AI models picked which of two product reviews to cite first. The one stating a price won in all six, and specs mattered in five (Vishwakarma et al., 2026). That's a test of anonymised reviews, not shop pages, and Sprinklr sells marketing software.

In a mock marketplace built from Amazon listings, a 10% price rise cut a product's odds of being picked by about 20% (Allouah et al., 2025). That held for Claude Sonnet 4, GPT-4.1 and Gemini 2.5 Flash under the default prompt. It's a preprint with MyCustomAI co-authors, and every listing showed its price.

An agent can't weigh a price it can't read.

Can AI read a price that JavaScript adds?

Not every system can. In searchVIU's October 2025 test, ChatGPT and Claude didn't find a JavaScript-inserted price on a live fetch. Gemini did, and so did Perplexity and Google AI Mode after indexing. None of the five found a price that existed only in JSON-LD (searchVIU, 2025). searchVIU sells a rendering checker.

To a crawler that doesn't run JavaScript, that page is a Bond Street jeweller's window with "price on application" on every tray.

Merchant Center help says to avoid "Using JavaScript or other animations to display prices" (Merchant Center Help, n.d.). Google says browser agents do render pages, reading screenshots, the DOM and the accessibility tree (Google, 2026).

A price in the server-rendered HTML works for both.

How do you find missing product facts with Rubric?

Run an audit in Rubric, the AI citability auditor we built, and open the Agent-ready tab. The tab counts pages with a machine-readable Offer, an explicit price and availability. It's advisory and doesn't change your score. Then check "price hidden until JS" and "stale or mismatched schema" on the Content citability signals card. Rubric predicts citability, it doesn't count citations.

What does the Agent-ready tab show on a service site?

On 5 October 2026, Rubric's Agent-ready tab found a machine-readable Offer on 15 of gogochimp.com's 182 crawled pages, an explicit price on 14 and availability on 7. We sell services, and most of the other pages are articles that shouldn't carry an Offer.

Click a signal to see its pages. On a shop, any product URL missing from the Offer, price or availability counts is your first fix. That's what Google's and OpenAI's docs ask for.

How should you read the price and schema signals?

Treat a clean "price hidden until JS" result as a first check, then search one product page's source for the price yourself. "Stale or mismatched schema" points to markup that doesn't match the page. Google's Merchant Center rule is that "Structured data must match the values that are shown to the customer" (Merchant Center Help, n.d.).

How do you check product pages with Claude and the Rubric MCP?

Connect Rubric's hosted MCP to Claude, pull the audit with get_audit, then run get_page on each product URL. Rubric's MCP page says get_page returns "one page's score, per-engine readiness, and every check result". Have Claude compare those results with the live page and draft what's missing. Then have someone who knows the stock approve each fact.

Create a key on the same page, which also gives you the Claude Code command with your key filled in. Reads are free, and crawls and draft checks run a live audit of your site.

Paste this into Claude:

Use list_audits to find my latest Rubric audit for example.com, then get_audit. Run get_page on each product URL below and list the checks it fails. Then read each live product page, compare it with those results, and draft the facts it's missing: price, currency, availability, brand, who the product is for, who it isn't for, and a spec table. Mark every guess for me to approve. Don't invent figures.
PASTE PRODUCT URLS HERE

/mcp__rubric__help lists the tools, and /mcp__rubric__fix runs Rubric's side of the loop.

How do you fix a product page so AI can read it?

Put the price, currency and availability in the HTML your server sends and in Product and Offer markup, and make them match. Then say who the product is for and who it isn't for, add a spec table, and keep any feed in step with the page.

What should every product page state?

Price, currency and stock status, plus what a good shop assistant says unprompted: brand, delivery, returns and who it suits.

Microsoft's Bing team gives an example of the "who it's for" line: "42 dB dishwasher designed for open-concept kitchens" instead of "quiet dishwasher" (Madhavan, 2025). No study I've found measures the "isn't for" line. I add it because it answers the buyer's next question, and Profound advises making "relevant limitations" easy to find (Profound, 2026).

Shirley Chen, Senior Director of Product Marketing for AI Commerce at Microsoft, puts the cost bluntly: "If your data is incomplete or out of date, you're not just lower in the results; you might eliminate yourself from the recommendation entirely" (Chen, 2026). It's vendor marketing, so read it as advice.

How do you fix it on Shopify?

Check that your product template outputs {{ product | structured_data }} in a JSON-LD script tag. Shopify's example output carries brand, price, currency and availability, as a ProductGroup when there are variants (Shopify, n.d.). If an app adds a second Product block, check the two don't disagree on price. Our Shopify CRO apps round-up covers the app side.

Shopify Catalog also sends your products to AI channels, and OpenAI says "No additional work is required from individual merchants" for ChatGPT (OpenAI, 2026). If titles or descriptions live in metafields, set up Catalog Mapping (Shopify, n.d.).

How do you fix it on WooCommerce and WordPress?

WooCommerce prints its Product JSON-LD server-side, with price, currency and stock status (WooCommerce, 2026). Core output doesn't include a brand, so assign one through the bundled Brands feature.

Variable products with differing prices get an AggregateOffer, and Google says "merchant listings require an Offer" (Google, 2026). My reading is that those pages won't reach merchant listings without a Merchant Center feed.

How do you fix it on Webflow Ecommerce?

Webflow's Schema markup field puts JSON-LD in the page head, and Webflow AI can pull Collection fields "like name, address, price, image" into it (Webflow, n.d.). Webflow doesn't say whether Ecommerce price and SKU fields can be bound. Test it on your own store, publish, and view the source.

What does a readable product page look like?

Like PLZ Soccer's Shopify product pages. We fetched three of them without JavaScript on 5 October 2026. Each carried a ProductGroup with the brand and an Offer at 45.00 or 50.00, in GBP and InStock. The visible "£45.00" was in the raw HTML too.

In September we rewrote the copy on all 89 product pages, taking the median from 135 to 381 words. The pages we sampled now restate the commercial facts in prose: "It costs £45, with free UK delivery over £50, 30-day returns and shipping to 41 countries" (PLZ Soccer, 2026).

Rubric scored the PLZ Soccer shop 75 on 12 September 2026 and 81 three re-crawls later that week, after work that included rewriting all 89 product pages. They're Rubric scores, not measured citations.

What's missing is reviews: none of the pages we sampled carries aggregateRating or review markup. Other client results, including Bee Friendly Skincare's Shopify page-speed work, are on our case studies page.

How do you confirm the fix worked?

Re-crawl with start_crawl and crawl_status, then run compare_audits to see which checks were fixed and which regressed. Run a product URL through Google's Rich Results Test too. A crawl can't tell you whether an answer changed, so ask the engines yourself.

Ask ChatGPT, Claude, Copilot and Google AI Mode a buyer's question that names a budget and a use. Log the date, engine and every shop named, and repeat over several days, as in our long-tail AEO guide. ChatGPT Shopping is US-only, so a UK test won't show what US shoppers see.

Whether the new copy sells more needs an A/B test, which our CRO methodology covers.

What do store owners ask about AI shopping and product markup?

Does Product schema get my pages cited by AI?

Not on its own. In Ahrefs' matched study of 1,885 pages that added JSON-LD, ChatGPT citations moved +2.2% and AI Mode +2.4%, both "statistically indistinguishable from zero". AI Overviews fell 4.6% (Linehan, 2026). Product markup still helps keep Google's shopping listings accurate.

Do AI agents need potentialAction or BuyAction markup?

I haven't found evidence they do. Google's web.dev guide says agents view sites through "screenshots, raw HTML, and the accessibility tree" (Kulikowski, 2026). OpenAI's feed spec and Microsoft's checkout pages don't mention potentialAction. Treat it as optional and unproven.

How do you keep a product feed and the page in step?

Change the price once and let it reach both. Merchant Center can use page markup to correct a feed: in Google's example, a product uploaded at $4 but shown at $3 becomes $3 (Merchant Center Help, n.d.). Google says this may fail if price or stock changes "more than once per day".

Do reviews and ratings affect AI product recommendations?

Ratings and reviews count when they're there. In the Allouah mock marketplace, all three models gave "positive weights to ratings and the number of reviews" (Allouah et al., 2025). OpenAI says ChatGPT "may consider available options, price, reviews, and ease of use" (OpenAI, 2026). Mark up only the reviews shown on the page.

What is the Universal Commerce Protocol?

A checkout protocol Google announced in January 2026 for product listings in AI Mode and the Gemini app (Srinivasan, 2026). Microsoft says feeds "primarily enable discovery", while UCP "enables transactions" (Microsoft, n.d.). A store publishes its profile at /.well-known/ucp (UCP, n.d.), and PLZ's Shopify store already serves one (checked 5 October 2026).

Should a service business add Offer markup?

Not for Google's shopping features. Google says "Only pages where a shopper can purchase a product are eligible for merchant listing experiences" (Google, 2026). That's why I'm not adding Offers to our blog posts. Mark up services where you publish a price.

Check one product page's source before you change anything

Pick a product that sells well, view its page source and search for the price. If the price, currency or stock status isn't in the HTML, fix the product template, because every product page shares it. If it's all there, write the who-it's-for line and the spec table. Our ecommerce conversion rate optimisation guide covers the human side of the same page. To check every product page in one go, run a Rubric audit.

References

• Allouah, A., Besbes, O., Figueroa, J. D., Kanoria, Y., & Kumar, A. (2025, v3 17 December). What Is Your AI Agent Buying? Evaluation, Biases, Model Dependence, & Emerging Implications for Agentic E-Commerce. arXiv preprint. https://arxiv.org/abs/2508.02630
• Chen, S. (2026, 21 August). How businesses win when AI does the shopping. Microsoft Advertising blog. https://about.ads.microsoft.com/en/blog/post/august-2026/how-businesses-win-when-ai-does-the-shopping
• GoGoChimp. (2026, 5 October). Rubric report for gogochimp.com, hosted crawl of 182 pages: Agent-ready tab. First-party.
• GoGoChimp. (2026, 5 October). Raw-HTML fetch, without JavaScript, of three PLZ Soccer product pages and of shop.plzsoccer.com/.well-known/ucp. First-party, unpublished.
• GoGoChimp. (2026, September). PLZ Soccer shop Rubric case study: work log for 12 to 14 September 2026 and Rubric scores. First-party, unpublished.
• GoGoChimp. (n.d.). Rubric MCP page, get_page tool description (wording as of 5 October 2026). Rubric's MCP page
• Google. (2025, updated 10 December). Introduction to Product structured data. Google Search Central. https://developers.google.com/search/docs/appearance/structured-data/product
• Google. (2026, updated 8 September). Merchant listing (Product, Offer) structured data. Google Search Central. https://developers.google.com/search/docs/appearance/structured-data/merchant-listing
• Google. (2026, updated 10 July). Optimizing your website for generative AI features on Google Search. Google Search Central. https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
• Google Merchant Center Help. (n.d.). About structured data markup for Merchant Center. https://support.google.com/merchants/answer/6069143?hl=en
• Google Merchant Center Help. (n.d.). About the Universal Commerce Protocol (UCP) and UCP-powered checkout feature on Google. https://support.google.com/merchants/answer/16837055?hl=en
• Google Merchant Center Help. (n.d.). Allow Merchant Center to update product information automatically. https://support.google.com/merchants/answer/12157888?hl=en
• Google Merchant Center Help. (n.d.). Free listings for products. https://support.google.com/merchants/answer/13889434?hl=en
• Google Merchant Center Help. (n.d.). Tips to help your products stay approved. https://support.google.com/merchants/answer/14999713?hl=en
• Kulikowski, K. (2026, updated 1 April). Build agent-friendly websites. web.dev. https://web.dev/articles/ai-agent-site-ux
• Linehan, L. (2026, 11 May). Ahrefs matched study of JSON-LD schema and AI citations (1,885 pages that added schema, 4,000 control pages). https://ahrefs.com/blog/schema-ai-citations/
• Madhavan, K. (2025, 8 October). Optimizing your content for inclusion in AI search answers. Microsoft Advertising blog. https://about.ads.microsoft.com/en/blog/post/october-2025/optimizing-your-content-for-inclusion-in-ai-search-answers
• Microsoft Advertising. (n.d.). Agentic Commerce (solutions page and FAQ). https://about.ads.microsoft.com/en/solutions/technology/agentic-commerce
• OpenAI. (2026, updated 4 October). Shopping with ChatGPT Search. OpenAI Help Center. https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search
• OpenAI. (n.d.). Get started. OpenAI Commerce developer docs. https://developers.openai.com/commerce/guides/get-started
• OpenAI. (n.d.). Power product discovery in ChatGPT (merchants page and FAQ). https://chatgpt.com/merchants/
• OpenAI. (n.d.). Products (Commerce feed specification, file upload). https://developers.openai.com/commerce/specs/file-upload/products
• Pichai, S. (2026, 11 January). The AI platform shift and the opportunity ahead for retail (NRF 2026 remarks). Google blog. https://blog.google/company-news/inside-google/message-ceo/nrf-2026-remarks/
• PLZ Soccer. (2026). Lionel Messi poster, product page copy as read on 5 October 2026. https://shop.plzsoccer.com/products/lionel-messi-poster
• Profound (Huang, A.). (2026, 2 October). The era of personal agents: Muse and Instinct. https://www.tryprofound.com/blog/the-era-of-personal-agents-muse-and-instinct
• Profound (Wu, A.). (2026, 3 September). ChatGPT 5.6 has transformed Shopping mode. https://www.tryprofound.com/blog/chatgpt-5.6-shopping-transformation
• Rockinson, R. (2023, 7 February). Google blog post explaining the Shopping Graph. https://blog.google/products-and-platforms/products/shopping/shopping-graph-explained/
• searchVIU. (2025, 2 December). searchVIU test of what ChatGPT, Claude, Perplexity and Gemini see in schema markup and price placements (test run 30 October 2025). https://www.searchviu.com/en/schema-markup-and-ai-in-2025-what-chatgpt-claude-perplexity-gemini-really-see/
• Shopify. (n.d.). structured_data Liquid filter. Shopify.dev. https://shopify.dev/docs/api/liquid/filters/structured_data
• Shopify Help Center. (n.d.). Selling on ChatGPT with agentic storefronts. https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/chatgpt
• Shopify Help Center. (n.d.). Shopify Catalog and product discovery for agentic storefronts. https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/products
• Southern, M. G. (2026, 9 September). Search Engine Journal report on ChatGPT Shopping results and product feeds. https://www.searchenginejournal.com/chatgpt-shopping-results-lean-hard-on-product-feeds/589000/
• Srinivasan, V. (2026, 11 January). New tech and tools for retailers to succeed in an agentic shopping era. Google blog. https://blog.google/products/ads-commerce/agentic-commerce-ai-tools-protocol-retailers-platforms/
• Universal Commerce Protocol. (n.d.). Specification overview (current release). https://ucp.dev/latest/specification/overview/
• Vishwakarma, R., Kumar, S., & Jamidar, R. (2026, 25 May). What Gets Cited: Competitive GEO in AI Answer Engines. SIGIR '26. https://arxiv.org/abs/2605.25517
• Webflow. (n.d.; "Updated 8 months ago" on 5 October 2026). Add schema markup in Webflow to improve SEO and AEO. Webflow Help Center. https://help.webflow.com/hc/en-us/articles/45766917223955-Add-schema-markup-in-Webflow-to-improve-SEO-and-AEO
• WooCommerce. (2026, trunk, last commit 30 September). class-wc-structured-data.php (Product JSON-LD output), with class-wc-brands.php for brand output. GitHub. https://raw.githubusercontent.com/woocommerce/woocommerce/trunk/plugins/woocommerce/includes/class-wc-structured-data.php