Category and product pages, now that AI answers do some of the shopping.

Shopping searches increasingly end in a summary rather than a list of links. Google shows product panels, AI Overviews and AI Mode responses that compare options; ChatGPT and other assistants return product cards with prices, ratings and merchants. Behind those answers is the same raw material: structured product data, merchant feeds, reviews and the pages that explain the products. That changes what an ecommerce page is for. It is still a landing page for a shopper, but it is also a data source for a system deciding which products to mention. This post is about the two templates that matter most for that, category pages and product pages, and how to build them so the page, the markup and the feed all tell the same story.

What changed, and what did not

The mechanics of ranking a category page in classic results have not changed much: relevant, crawlable, well linked, useful. What has changed is how much of the shopping journey now happens in a generated answer before the click.

None of the major platforms fully document how they select products for these answers, and anyone claiming to know the exact formula is guessing. What is documented is the input. Google says it uses Merchant Center data and structured data on your pages to understand products, and its merchant listing documentation describes the price, availability, shipping and returns information it reads. Assistants that browse the web read the same pages and the same review sources everyone else can.

So the practical question is not how to game an AI shopping answer. It is whether your pages give a clear, consistent, specific account of what you sell, in a form both people and machines can use. We cover the wider picture, including the marketplace problem, in AI search for ecommerce. This post stays on the two templates.

Category pages answer the 'which one' question

Someone searching “trail running shoes for wide feet” or asking an assistant the same thing is not looking for a specific product yet. They are trying to narrow a choice. A category page that is only a grid of products does not help with that, and gives a summarizing system nothing to work with either.

What we add to category templates, in order of value:

• A short, specific intro above the grid. Two to four sentences that define what is in the range and name the decision that matters most (drop, cushioning, terrain). Not keyword copy.
• A guide to the sub-types. A small table or a few lines explaining the difference between the main options, each linking to its subcategory. This is often the most quoted part of a category page.
• Filters named in shopper language. If customers say “wide fit”, the filter should say “wide fit”, not “Width: E/2E”. Filter labels are content.
• Buying guidance below the grid. Sizing, compatibility or care, where it genuinely helps. Keep the products first for people who already know what they want.
• Links to related categories and guides. So both crawlers and shoppers can move sideways.

What we remove: long blocks of text written for search engines, hidden behind “read more” at the bottom of the page, repeating the category name. They do not help shoppers and they do not tend to be what anyone quotes.

Facets: which filtered pages deserve to exist

Faceted navigation is where ecommerce sites create most of their crawl waste. Every combination of size, color, brand, price and sort order can become a URL. Google’s own guidance on faceted navigation is blunt about the cost: crawling huge numbers of filter URLs can slow the discovery of pages that matter.

The approach we use is demand-led:

• Find the filter combinations people search for. “Waterproof hiking boots” or “leather sofa beds” are real queries. Most combinations (“blue, size 9, under $80, sorted by newest”) are not.
• Give the ones with demand a proper landing page. A static, crawlable URL with its own title, intro and internal links, rather than relying on a parameter URL to rank.
• Keep the rest out of the crawl. Using robots rules, consistent parameter handling and internal links that do not point at every combination. Canonical tags alone are a hint, not a crawl control.
• Never let sort orders or session parameters create indexable pages.

This also makes category pages better material for AI answers. A page called “waterproof hiking boots” with a clear intro and a curated range is a far more useful source than a parameter URL with a product grid and nothing else.

Product pages answer 'is this the right one'

By the time someone lands on a product page, or asks an assistant about a specific product, they want confirmation. The pages that do well in this job are specific in ways manufacturer copy never is.

• Specifications as a table, not a paragraph. Dimensions, materials, weight, compatibility, what is in the box. Structured specs are easy to read, easy to compare and easy to extract.
• Who it suits and who it does not. “Best for wide feet on mixed terrain; not built for racing.” Honest fit statements are exactly the kind of sentence that answers a shopping question.
• Comparison with sibling products. If you sell three similar models, say how they differ. Otherwise the comparison gets made somewhere else, by someone else.
• Sizing, fit and compatibility guidance. The questions your support team answers every week belong on the page.
• Shipping and returns, stated plainly. They influence the decision, and Google’s merchant listing data supports them directly.

You do not need to rewrite a whole catalog at once. Start with the products that sell most and the ones with search demand, and add what only you know.

Make the page, the markup and the feed agree

This is the part most stores get wrong, and it is the least glamorous. The same product is described in three places: the visible page, the structured data on that page, and the product feed you send to Google Merchant Center (and, increasingly, to other shopping surfaces). When they disagree, every system downstream has to decide which to believe.

• Price and availability must match across page, markup and feed. Merchant Center checks for mismatches and can limit products that fail.
• Identifiers (GTIN, MPN, brand) should be present wherever they exist. They are how a system knows your listing and a marketplace listing are the same item.
• Variants should be represented consistently, with each variant’s price, size and availability correct in both markup and feed.
• One Product block per page. Themes, review apps and SEO plugins often each add their own. Two blocks that disagree are worse than one that is complete.
• Shipping and return policies set up in Merchant Center or marked up on the page, and kept in step with what your policy page actually says.

Merchant Center is not only for ads. Google uses product data from it for free listings across its shopping surfaces, which is why we treat feed quality as an organic search task rather than something that belongs to the paid team.

Reviews and questions are content too

Review text is one of the richest sources of the language shoppers use, and it is widely read by systems summarizing products. Volume helps, but substance helps more: a review that says how a product held up after six months of use is more useful than five stars and “great!”.

• Ask for reviews in a way that prompts detail: what did you use it for, how does it fit, would you buy it again.
• Show reviews on the product page in crawlable HTML, not only inside a script widget.
• Answer product questions publicly, and fold the common ones back into the description.
• Do not incentivize positive reviews or suppress negative ones. Beyond being a bad idea, the FTC’s rule on fake reviews and testimonials makes some of these practices unlawful in the US.

What we would not do

There is a lot of advice circulating about winning AI shopping answers. Some of it is harmful.

• Mass-generated product descriptions. Rewriting thousands of descriptions with a language model and publishing them unedited is the kind of scaled content Google’s spam policies target. We saw what that looked like in the August 2026 spam update.
• A “best X for Y” landing page for every adjective. Pages created only to catch query variations, with the same products in a different order, read as doorway pages.
• FAQ markup in the hope of rich results. Google limits FAQ rich results to a small set of authoritative sites. Write FAQs for shoppers; do not expect a snippet.
• Claims you cannot support. “Best-selling”, “#1 rated”, “clinically proven”. Assistants repeat what you say, which makes accuracy more important, not less.

How to tell whether it is working

Measure the classic and the new surfaces separately, because they move at different speeds.

• Search Console: category and product template performance, plus the merchant listing and product snippet reports for structured data errors.
• Merchant Center: product disapprovals, mismatches and free listing performance.
• A fixed set of shopping prompts: the category questions your customers actually ask, run regularly in Google AI Mode and ChatGPT, logging which products and merchants are named.
• Revenue by landing template: because a category page that ranks but does not sell has not done its job.

We wrote more about building that prompt set and reporting it honestly in measuring AEO ROI. If you want this done across your catalog, our ecommerce SEO team does exactly that work.

Frequently asked questions

How do I get my products into ChatGPT shopping results?

There is no guaranteed method, and the selection process is not fully documented. Make sure your product pages are crawlable, specific and consistent, keep structured data and feeds accurate, earn detailed reviews, and check each platform's current merchant documentation for any feed or program it offers.

Does Google AI Mode use Merchant Center data?

Google says it uses Merchant Center product data and on-page structured data to understand products across its shopping experiences. Keeping your feed accurate and matching your pages is the most reliable thing you control.

How much text should a category page have?

Enough to help someone choose: a short intro above the grid, a guide to the main sub-types and useful buying guidance below. Long keyword-heavy blocks at the bottom of the page rarely help shoppers or rankings.

Should filtered category pages be indexed?

Only the filter combinations with real search demand, ideally as dedicated landing pages with their own content. Keep sort orders and low-value combinations out of the crawl.

Should out-of-stock products stay on the site?

If they are coming back, yes: keep the page, mark availability accurately in the markup and feed, and point to alternatives. If they are gone for good, redirect to a replacement or the closest category where it helps shoppers.

Do I need to rewrite manufacturer product descriptions?

For your most important products, yes: add fit, use cases, comparisons and answers to common questions. You do not need to rewrite the whole catalog at once, and mass AI rewrites published unedited are a risk.