E-commerce SEO for AI search: product and category pages that Google and ChatGPT understand

E-commerce SEO is the work of making sure Google and AI assistants can find, understand and recommend your product and category pages. It rests on four things: a catalogue that can be crawled without getting lost in duplicates, product pages with original descriptions, product data marked up to match the page, and a feed in Google Merchant Center.

To check it on your own site: Online shops

An online shop plays by the same rules as any website, with three problems a service business never has: hundreds or thousands of near-identical pages, data that changes daily (price, stock) and filters that multiply URLs. This guide brings together what Google Search Central, schema.org and OpenAI document about each of them, without numbers of doubtful origin or tricks nobody has shown to work.

What is different about SEO for an online shop?

Google gives e-commerce its own section in Search Central, and opens it with the core problem: a critical challenge for any e-commerce site is being discovered in Search1. Its answer has two parts: share your product data with Google, and build a site structure Google can understand1.

Each type of page in a shop answers a different kind of search and needs different work:

How do you get a product page to rank?

A product page is the page where a specific item can be bought: it carries the item's name, photos, price and the add-to-basket button. It is the page Google can show with a price and star rating, and the one an AI assistant links to when it recommends a product.

Write your own description, not the manufacturer's

The manufacturer's description is used by every shop that stocks the product. Google has no reason to prefer your copy over fifty others. Its self-assessment questions put it plainly: if content draws on other sources, does it avoid simply copying or rewriting them, and add substantial value and originality?2

An original description does not mean a longer one. It means saying what only the seller knows:

• Who it suits and who it does not. "Narrow fit: if you take a wide 5, order a 5.5."
• The measurements that matter when buying, in a table: weight, dimensions, materials, compatibility.
• What customers ask in the shop or by email, answered on the page itself.
• Your own photos of the product in use, alongside the catalogue shots.

Answer the buying questions on the page

A shopper decides with five facts: what it is, what it costs, whether it is in stock, how long delivery takes and how returns work. If any of them only appears in the basket or on a legal page linked from the footer, it is harder for Google or an assistant to connect it with that product.

Write them on the page, as text: "Next-day delivery on orders placed before 2pm, free over £50. Free returns within 30 days." For its AI features, Google asks that important content is available in textual form3. How to write sentences an assistant can lift and quote is covered in the guide on how to get cited by AI.

Variants: one URL that preselects each one

Sizes and colours are a classic source of duplicates. Google recommends a distinct URL for each variant, using a path segment (/t-shirt/green) or a parameter (/t-shirt?color=green), and warns against fragments: to Google, /t-shirt#black and /t-shirt#white are the same page4.

For the markup, Google documents the ProductGroup type. It groups the variants of a product with hasVariant, says what they vary by with variesBy (size, colour) and identifies the parent product with productGroupID5. Two set-ups are supported: all variants on a single page with one canonical URL for the group, or each variant on its own page with full, self-contained markup5.

What structured data does an online shop need?

Product structured data is a block of code, usually JSON-LD, that states unambiguously which product a page sells, at what price and on what terms. It uses the schema.org vocabulary, which defines Product as any offered product or service6 and Offer as an offer to transfer some rights to an item, such as selling it7.

Google distinguishes two uses of product markup8:

• Product snippets, for pages where people cannot buy the product directly, such as an editorial review.
• Merchant listings, for pages where customers can buy the product from you. This is the one that matters to a shop.

Required and recommended properties

This table summarises Google's documentation for merchant listings9:

Two details that save mistakes. Only pages where a shopper can buy the product are eligible for merchant listings, not pages that link to another site that sells it9. And Google recommends declaring a store-wide returns policy and shipping policy once, under your Organization markup, rather than repeating them on every product9. If you also configure shipping in Merchant Center or Search Console, Google gives those settings precedence over the markup on your pages10.

An example of marked-up product data

This example is for a fictional shop: example.com is a domain reserved for examples, and the product, price and ratings are made up. Shipping and returns are left out to keep it short.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Example Ridge women's trail shoe",
  "image": [
    "https://www.example.com/img/ridge-womens-side.jpg",
    "https://www.example.com/img/ridge-womens-sole.jpg"
  ],
  "description": "Waterproof trail shoe with a narrow fit, built for mid-distance mountain routes.",
  "sku": "RIDGE-W-5-BL",
  "brand": { "@type": "Brand", "name": "Example Outdoor" },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": 4.6,
    "reviewCount": 38
  },
  "offers": {
    "@type": "Offer",
    "url": "https://www.example.com/trail-shoes/ridge-womens?size=5&colour=blue",
    "price": "109.00",
    "priceCurrency": "GBP",
    "availability": "https://schema.org/InStock",
    "itemCondition": "https://schema.org/NewCondition"
  }
}

Three rules so the markup helps rather than hurts:

1. What you mark up has to be on the page. Google asks for structured data that matches the visible text for its AI features3. If the markup says £109 and the page says £89 because of a sale, the markup is wrong.
2. Ratings are the ones shoppers can see. The ratingValue and reviewCount must be the average and count shown on the page, not a rounded figure.
3. Generate it from your platform, not by hand. Shopify, WooCommerce or PrestaShop build the markup from catalogue data, natively or with an extension. Hand-written JSON goes stale with the first price change.

How do you get a category page to rank?

A category page is the listing that groups products of one type: "trail running shoes", "stovetop espresso makers". It answers broader searches than a product page, usually with more volume, because the shopper is still comparing.

Google explains that the more internal links a page receives, the more important it looks relative to the rest of the site. It suggests a clear hierarchy: links from the menus to the categories, from categories to subcategories, and from subcategories to every product11. To push a particular product, link to it from the home page or from other content such as blog posts11.

What most often goes wrong on shop categories:

• Listings with no copy. A grid of photos does not explain what you sell or help anyone choose. Two or three useful paragraphs ("how a trail shoe differs from a road shoe, how to pick your size") give Google and assistants something to understand and quote.
• Links that are not links. Google asks you to use <a href> elements, not JavaScript events on other elements11. A product that can only be reached by clicking a div may never be discovered.
• Pagination that only works with a button. Google's crawlers do not click buttons and generally do not trigger JavaScript functions that need a user action12. A "Load more" button with no links behind it hides every product from page two onwards.

Pagination Google can follow

Google gives four rules for paginated listings12:

1. Link each page to the next with <a href> tags.
2. Give each page its own URL, for example with ?page=2.
3. Do not use the first page as the canonical for all of them: each page carries its own canonical URL.
4. Do not use # fragments for page numbers.

On the third: if page 2 of a category points its canonical at page 1, you are telling Google that page 2's products are on page 1, and they are not.

Filters, duplicates and canonicals: keeping Google out of the maze

Faceted navigation means the filters on a listing (size, colour, brand, price, rating) when every combination creates its own URL. With five filters of ten options each, a single category can produce tens of thousands of URLs showing almost the same products.

Google describes two concrete problems with those URLs13. The first is overcrawling: each combination looks like a new URL, so the crawler visits a great many before realising they are useless. The second is slower discovery: time spent on filters is time not spent on your new products.

Which tool to use for which case

The fix depends on whether you want those filter URLs to appear in Google. This table brings together what Google documents for each situation:

Three mistakes to avoid:

• Using noindex to save crawl budget. Google still requests the page to read the noindex, then drops it, so the crawling time is spent anyway15.
• Combining noindex and Disallow on the same URL. For noindex to work, the page must not be blocked in robots.txt; if it is, Google never sees the tag16.
• Using robots.txt or noindex to choose a canonical. Google advises against it: noindex removes the page from Search entirely, and a URL blocked in robots.txt can still be indexed14.

One last point: rel="canonical" is a hint, not a rule. Google may choose a different page as canonical from the one you indicate17. So make internal links, the sitemap and the canonical all say the same thing, and add a self-referencing canonical on every indexable page, as Google recommends4.

Does your shop need to worry about crawl budget?

Crawl budget is the time and resources Google spends crawling a site. Google publishes a guide to managing it, but is clear about who it is for: sites with over a million unique pages that change about once a week, sites with over ten thousand pages that change daily, or sites with many URLs reported as "Discovered – currently not indexed"15.

If your shop does not have a large number of rapidly changing pages and new products are crawled on the day they go live, Google says keeping your sitemap up to date and checking the Page indexing report regularly is enough15.

What does affect shops of any size is filters, because they turn a catalogue of a few thousand products into a site with hundreds of thousands of URLs. Beyond the measures above, Google's guide adds keeping sitemaps current, avoiding long redirect chains and making pages efficient to load15. And if a product is not linked from any category, Google suggests listing it in a sitemap or a Merchant Center feed so it can be found11.

Speed on listings and product pages: Core Web Vitals

Core Web Vitals are three user-experience metrics (loading, interactivity and visual stability) that Google's ranking systems use18. The "good" thresholds are an LCP (when the main content renders) of 2.5 seconds or less, an INP (response to an interaction) of 200 milliseconds or less and a CLS (layout shift) of 0.1 or less, measured at the 75th percentile of page loads, separately for mobile and desktop19.

Keep them in proportion. Google says its Search always seeks to show the most relevant content, even if the page experience is sub-par18. A fast product page with a copied description will not beat a slightly slower one that answers the shopper better. Speed breaks ties; it does not replace content.

Three performance problems that come up again and again on listings:

Measure the template, not a single page. If the category template has an LCP problem, every category has it, and fixing it once fixes them all.

Reviews and ratings: what counts

Reviews serve two audiences: the hesitant shopper and the systems deciding what to show. To mark them up, Google requires an average rating (ratingValue) in AggregateRating plus at least the number of ratings (ratingCount) or reviews (reviewCount)22.

Google's conditions for the stars to count22:

• Marked-up reviews must be visible on the marked-up page, and it must be obvious to users that the page contains reviews.
• Ratings must come directly from users, with no fake or undisclosed incentivised reviews.
• Don't aggregate reviews or ratings from other websites.
• LocalBusiness and Organization pages are not eligible for stars when the entity controls the reviews about itself.

The practical upshot for a shop: product reviews belong on the product page, marked up; reviews of the shop as a whole belong on external platforms. Beyond on-page reviews, a shop can publish its own buying guides and comparisons. For that content, Google recommends showing expertise, providing evidence of first-hand experience (your own photos, measurements), explaining what sets a product apart from competitors and discussing pros and cons based on your own research23. It adds that what counts is quality and originality, not length23.

Google Merchant Center and free listings

Google Merchant Center is Google's free tool where a shop uploads its catalogue as a product feed. With free listings, those products can appear at no cost on Google Search, Google Maps, Gemini, YouTube, the Shopping tab, Google Images and Google Lens24.

It is not required to appear in ordinary search results. Google says submitting data through Merchant Center is not mandatory to appear in Google Search results, but it can improve Google's understanding of your products, and a feed gives you more control over the timing of updates25. It is, however, mandatory for some surfaces, such as the Shopping tab25. In its product markup documentation, Google adds that providing both structured data on your pages and a Merchant Center feed maximises your eligibility for shopping experiences8.

The rule that ties the two together is the same as for markup: the feed, the markup and the page must say the same thing. Google also asks you to keep Merchant Center information up to date for its AI features3.

How do AI assistants decide which products to recommend?

No assistant publishes its recommendation algorithm, but Google and OpenAI document enough to know what data they use. For the wider picture, beyond shops, the method in the AI visibility audit guide applies.

Google: AI Overviews, AI Mode and Gemini

Google says there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode3. You do not need new AI text files or any special schema.org markup3. Its recommendations are standard SEO, and three of them are pure e-commerce:

• Make sure your structured data matches the visible text on the page.
• Keep your Merchant Center information up to date.
• Put important content in text form, not only in images.

Gemini is also listed among the places where Merchant Center free listings can appear24. For a shop, an up-to-date feed is one of the documented routes into the products Google's AI shows; markup that matches the page is the other.

ChatGPT

OpenAI's help centre explains how shopping in ChatGPT works26:

• Products are selected independently: they are not ads and are not influenced by OpenAI partnerships. A product appears when ChatGPT judges it relevant to what the user asks.
• It considers structured metadata from first-party and third-party providers, such as price and description, as well as other third-party content.
• When several merchants sell the same item, it ranks them by availability, price, quality and whether they are the maker or primary seller.
• Shopify shops are already integrated through Shopify Catalog, with no extra work. Other merchants can apply to send OpenAI a product feed directly.

A caveat on that direct feed if you sell outside the US: OpenAI's specification says the standard upload currently targets the US, that leaving the country empty does not mean worldwide, and that multi-country support has to be confirmed for each integration27. Check where you stand before relying on that route.

The route that needs no agreement is the open web. OpenAI recommends allowing OAI-SearchBot in robots.txt, and accepting requests from its published IP ranges, so your site can appear in ChatGPT search28. GPTBot is a different crawler, used for training, and OpenAI says each setting is independent: you can disallow GPTBot and still allow OAI-SearchBot28.

What this means for your shop

Both sets of documentation point the same way, though this is an inference rather than a guarantee: complete product data that matches across the page, the markup and the feed; reviews beyond your own site, because ChatGPT takes third-party content into account; and search crawlers given free passage in robots.txt and through your firewall.

How do you know whether AI sends you visitors?

Visits from assistants show up in Google Analytics 4, which added a default AI Assistant channel in May 2026 for traffic arriving from sources such as ChatGPT, Gemini or Copilot29. AI Overviews and AI Mode are not in it: GA4 counts them as organic search29. Search Console includes them in web search traffic3 and, since June 2026, in a dedicated report that for now shows impressions only30.

A mention without a click leaves no trace in either tool. That is why, in a client report, AI visibility and AI traffic are best kept as separate lines; the SEO client report guide covers how to present them.

E-commerce SEO checklist

1. Every product page has an original description, not the manufacturer's copy as-is.
2. Price, stock, delivery time and returns policy appear on the product page as text.
3. Every product page carries Product markup with an Offer: name, image, price and currency, plus availability, shipping, brand and ratings if you have them.
4. The markup matches what is visible: same price, same stock, same average rating.
5. Each variant has a URL that preselects it, with no #, and variants are grouped with ProductGroup or a clear canonical.
6. Category pages have useful copy and link to their products with <a href>.
7. Pagination uses real links, one URL per page and its own canonical on each page.
8. Filters you do not want indexed are blocked in robots.txt, with no noindex on those same URLs.
9. Permanently withdrawn products return 404 or 410.
10. The sitemap is current and the Page indexing report does not show thousands of filter URLs.
11. Images on the first screen are not lazy-loaded, and every image has dimensions.
12. The catalogue is in Google Merchant Center with free listings switched on, and the feed matches the site.
13. robots.txt lets Googlebot and OAI-SearchBot in.

Declaration of interest: this blog belongs to SmoothSeen, a web audit tool that measures visibility in search engines and AI assistants. It analyses one page at a time, product or category. On a product page it checks whether the product markup declares price, availability, brand, shipping, returns and ratings, whether there is an original description and an FAQ, and measures speed with Core Web Vitals. It delivers a search visibility score and an AI visibility score, a comparison with up to three competing shops, a prioritised action plan and a PDF report, under your brand on white-label plans. It does not track rankings, does not connect to your platform, Google Merchant Center, Search Console or Google Analytics, and does not analyse a whole catalogue at once. What it checks for shops is on the online shops page.

What to do this week

Pick your most important category and one of its best-selling products. On the product page, check that price, stock, delivery and returns are written out and marked up; on the category page, check that the filters are not creating indexable URLs unchecked. If you would like those checks measured alongside your competitors', analyse the product page with SmoothSeen.

Frequently asked questions

How do I get my online shop to rank on Google?

Start with whatever stops Google seeing your catalogue: crawlable links from the menu to the categories and on to every product, filters that do not create thousands of duplicates, and an up-to-date sitemap. Then work on product pages, with original descriptions and marked-up product data, and on categories, with copy that helps people choose. Speed and reviews come after that.

Can I use the manufacturer's description on my product pages?

You can, but it will not help you rank. It is the same text every shop selling the product publishes, and Google's quality questions ask whether content adds substantial value beyond copying or rewriting other sources. Use it as a technical starting point and add what only the seller knows: who it suits, how it fits, what customers ask, and your own photos.

Do I need Google Merchant Center to appear in Google Shopping?

For the Shopping tab and the other free listing surfaces, yes: free listings are what put your products in front of shoppers there at no cost. For ordinary search results it is not required, because Google can read the product markup on your site. Google recommends using both, because together they maximise your eligibility for shopping experiences.

How do my products get into ChatGPT?

ChatGPT chooses products using structured metadata from first-party and third-party providers, such as price and description, plus third-party content. Shopify shops are already integrated through Shopify Catalog; other merchants can apply for a direct feed, though the standard upload currently targets the US. The open route is to allow OAI-SearchBot in robots.txt and publish complete data on every product page.