Blog

Ecommerce SEO: from category pages to AI shopping

Online shops win in Google with strong category pages, complete product data and technical control over filters and variants.

Key takeaways

  • Category pages are the engine of ecommerce SEO: they rank for the broad keywords with buying intent.
  • Faceted navigation is the biggest technical pitfall, because filter combinations create an almost endless number of near-identical URLs.
  • Do not simply delete out-of-stock products: temporarily unavailable stays live, permanently gone gets a relevant 301 or a 404/410.
  • Product structured data and a Google Merchant Center feed with free listings belong to SEO, including visibility in AI answers.
  • AI shopping (AI Mode, ChatGPT, UCP) mainly demands complete, consistent product data; UCP launched in the US first and is something to watch rather than act on for most shops outside it.

Ecommerce SEO comes down to three things: category pages that match searches with buying intent, product pages that give shoppers and search engines every fact they need, and a technical foundation that stops filters and variants from generating thousands of worthless URLs. Get those three right and you are found in the regular results, in Google Shopping and increasingly in AI answers.

This guide is for owners and marketers of online shops. I am writing it in October 2026, a year after Google’s AI Mode launched in the Netherlands on 8 October 2025, while Google and OpenAI build shopping into AI assistants. The fundamentals of ecommerce SEO have not changed, but the bar for good product data is higher.

Why are category pages the engine of ecommerce SEO?

Category pages rank for the broad, commercial keywords where most of the search volume sits, such as ‘women’s running shoes’ or ‘aluminium garden furniture’. Product pages mostly rank for specific brand and model searches. Build the category structure well and you have laid the foundation for the whole shop.

A good category page is not a product grid with 800 words of SEO copy at the bottom that nobody reads. It is a page that helps the searcher choose:

  • a title and H1 that match the search term (‘Women’s running shoes’, not ‘Collection’);
  • a short, useful introduction above the products: who this category is for and what to look out for;
  • products and filters present in the HTML, so a crawler (the program search engines use to fetch pages) can read them without executing JavaScript;
  • links to subcategories and, where it makes sense, to a buying guide.

Let keyword research decide which categories and subcategories you create, not just the way your buying team organises stock. If lots of people search for ‘trail running shoes’, that deserves its own page, even if your ERP system treats it as a filter value. According to Google’s guide to generative AI features (May 2026) there is no ideal page length: write as much as the buyer needs.

How do you build a product page that ranks and sells?

A strong product page gives every fact a buyer is looking for, in copy that does not appear on a hundred other shops. Most retailers use the same manufacturer description as their competitors, which leaves Google little reason to show your page in particular.

  • Your own description covering use, how it differs from similar products and, where possible, hands-on experience. Google calls this ‘non-commodity’ content: information that is not the same everywhere.
  • Specifications in an HTML table: dimensions, material, weight, compatibility. These are exactly the facts AI search engines extract from pages.
  • Price, delivery time, shipping costs and returns policy visible on the page, not hidden until checkout.
  • Good images with descriptive file names and alt text, plus video where it genuinely helps.
  • Customer reviews and questions, covered below.

How do you stop faceted navigation from getting out of hand?

Faceted navigation is the biggest technical pitfall for online shops. Every combination of brand, colour, size, price and sort order can get its own URL, so a category with ten filters soon produces more URLs than you have products. Search engines then waste time on near-identical pages and find your important pages later.

Google has dedicated documentation on faceted navigation. The gist: if you do not need filter URLs in search results, prevent them from being crawled; if you do need them, keep the URLs consistent and return a 404 for combinations with no results. In practice I decide per filter type:

Filter typeExampleSearch demand?Recommendation
Brand within category/running-shoes/nike/OftenIndexable landing page with its own title, H1 and intro
Key attribute/running-shoes/trail/SometimesIndex if there is demonstrable search volume
Size, price range, stock?size=9&price=50-100RarelyDo not index, limit crawling
Sorting and display?sort=price-ascNeverKeep out of the crawl, canonical to the category
Combinations of three or more filters?brand=x&colour=y&size=zAlmost neverNo crawlable links, block in robots.txt if needed

Watch out for one classic mistake: Google cannot fetch a URL you block in robots.txt, so it will never see a noindex tag or canonical on that page. Choose one method per URL pattern. And a canonical is a hint, not an instruction: if hundreds of filter URLs are linked internally, Google may still crawl and occasionally index them.

After more than ten years of technical SEO at iProspect (Dentsu) and DEPT®, where I worked on large brands such as eBay and Philips (clients of those agencies), I start every ecommerce review in the same place: how do the filters behave? That is usually where crawl budget and indexing leak away, and it is rarely visible to anyone looking only at the front end of the shop.

How do you deal with duplicate content in an online shop?

Duplicate content in ecommerce is almost always caused by technology, not bad intent. You fix it by choosing one fixed URL per product and per category and pointing every other version at it. Google lists ‘less duplicate content’ as a technical point in its AI guide too.

  • Products in several categories: use a product URL without a category path (/product/name/) or one fixed canonical, so the same product does not live under three paths.
  • Variants: if people search for colours separately (‘red iPhone case’), separate pages can make sense. Otherwise use one page with a variant selector.
  • Pagination: keep page 2, 3 and beyond crawlable through normal links and give each page a self-referencing canonical, not one pointing to page 1.
  • Technical variants: http and https, with and without www, with and without a trailing slash: one version, the rest 301-redirected.

What should you do with out-of-stock products?

Do not simply take an out-of-stock product offline. If it is temporary, keep the page live with the correct availability; if it is permanent, redirect to a genuinely comparable alternative or let the page go cleanly.

  • Temporarily out of stock: the page stays live, availability is ‘OutOfStock’ in structured data and the product feed, with an expected date and alternatives visible.
  • Discontinued, with a successor: 301 redirect to the successor model.
  • Discontinued, no real alternative: a 404 or 410 is fine. Redirecting everything to the homepage or a broad category often backfires, because Google may treat such a redirect as a ‘soft 404’.
  • Still searched for or linked to: consider keeping it as an information page with the manual and successor.
  • Seasonal products: reuse the same URL every year so accumulated signals are not lost.

Large changes to your range are effectively a small migration; my article on website migration SEO explains how to build a redirect plan.

Which structured data does an online shop need?

An online shop needs at least Product markup with an Offer (price, currency, availability), plus review data and, where possible, shipping and returns information. Structured data is code in the page, usually JSON-LD, that tells search engines in a fixed format what the page contains.

According to Google’s Product structured data documentation there are two kinds of rich result: product snippets (such as stars and price on a regular result) and merchant listings for pages where the product can be bought directly. Use Product, Offer, AggregateRating and Review, and add shipping costs and returns policy.

Two sober caveats. First, Google says structured data is not required to appear in AI Overviews or AI Mode, and there is no special ‘AI markup’. It is, however, recommended for rich results. Second, price and availability in the markup must match exactly what is on the page and in your Merchant Center feed. More in my article on llms.txt, schema and AI crawlers.

Why is Google Merchant Center part of ecommerce SEO?

Google Merchant Center is part of SEO because, since 2020, your products can be shown free of charge through free listings, with no advertising budget. In its AI guide (May 2026) Google also names Merchant Center feeds as a way for products to appear in AI answers.

Many shops use Merchant Center only for ads. Here is how to set up the organic side too:

  1. Create a Merchant Center account and verify and claim your domain.
  2. Submit a product feed (through your shop platform or a plugin) with unique IDs, GTINs where available, correct prices and availability.
  3. Enable free listings and fill in your shipping and returns settings.
  4. Resolve disapproved products and warnings; every disapproval is a product that is not shown.
  5. Check monthly that the feed, structured data and the page still agree with each other.

Reviews add unique, current text that answers exactly the questions other buyers have, and they make star ratings in search results possible. Internal links determine which pages search engines see as important and how quickly new products are found.

Collect reviews at product level, render them in the HTML (not only in a widget that loads later) and mark them up with Review and AggregateRating. Never buy or invent reviews.

For internal linking:

  • use breadcrumbs with BreadcrumbList markup;
  • link from buying guides and blog posts to the relevant categories, with descriptive anchor text;
  • show related and complementary products as real links;
  • keep the main menu manageable: hundreds of links in a mega menu dilute the signals.

How fast does an online shop need to be?

An online shop should at least reach ‘good’ on all three Core Web Vitals: Largest Contentful Paint (LCP) within 2.5 seconds, Interaction to Next Paint (INP) under 200 milliseconds and Cumulative Layout Shift (CLS) under 0.1, according to web.dev. Speed is no magic ranking lever, but it does matter for conversion.

The usual culprits in shops are large hero images, filter scripts and, above all, third-party scripts: chat, review widgets, A/B testing tools and tracking.

I saw how much can be gained on my own site racemotion.nl. After moving from WordPress to static HTML in October 2026, its mobile PageSpeed Insights score went from 71 to 100 and LCP from 6.2 to 1.4 seconds. You cannot simply turn a shop into a static site, but the lesson is broader: anything you do not load cannot slow you down. If in doubt, have your templates and scripts reviewed as part of a technical SEO analysis.

What do AI Mode and ChatGPT shopping mean for your shop?

AI shopping mainly changes where customers come across your products, not what you need to do: complete, consistent product data and content that answers comparison questions.

Where things stand in October 2026:

  • Google AI Mode breaks a question such as ‘best running shoe for beginners with wide feet’ into several searches (query fan-out) and grounds the answer in pages from Google’s index. According to Google, Merchant Center feeds help products appear in those answers. Read more in my article on AI Mode and AI Overviews.
  • ChatGPT has shopping features that show product results. If you want to be visible there, do not block OAI-SearchBot in robots.txt; according to OpenAI that is the crawler for ChatGPT’s search features, separate from GPTBot, which is used for training.
  • Universal Commerce Protocol (UCP) is an open standard that lets AI agents make purchases on a customer’s behalf. Google launched it in January 2026, in the United States first. For most shops outside the US it is something to watch in October 2026, not something to invest in yet.

What not to do: create hundreds of pages for every conceivable question variant. Google warns in its AI guide that mass-producing pages for search variations or fan-out queries can count as ‘scaled content abuse’. One strong buying guide per product group is worth more. If you want to know how AI search engines currently present your brand, AI search optimisation is the logical next step.

My advice: where do you start?

Start with the technical issues that affect everything, then work from your most important categories down to your products. This is the order I use myself:

  • Crawl the shop and map how many URLs filters and sorting generate.
  • Decide per filter type: index, noindex or keep out of the crawl.
  • Resolve duplicate product and category URLs with one fixed URL and 301s.
  • Rewrite the title, H1 and intro of your twenty most important categories based on keyword research.
  • Add complete Product markup and check it in the Rich Results Test.
  • Connect Merchant Center and enable free listings.
  • Agree a fixed process for out-of-stock and discontinued products.
  • Measure Core Web Vitals on your category and product templates and remove unnecessary scripts.

If you want to know where your shop stands, it starts with an SEO audit that covers exactly these points. I do not guarantee rankings; I do give you a clear picture of what is holding you back.

Sources

  1. Google Search Central, Guide to Optimizing for Generative AI Features on Google Search (May 2026)
  2. Google Search Central, Managing crawling of faceted navigation URLs
  3. Google Search Central, Product structured data
  4. web.dev, Web Vitals
  5. OpenAI, Overview of OpenAI Crawlers

Frequently asked questions

How much text should a category page have?

There is no ideal word count; Google itself says there is no ideal page length. Put a short introduction above the products and place more detailed buying advice or FAQs further down the page. Copy written only for search engines helps nobody.

Do I need to rewrite every manufacturer product description?

Not all at once. Start with the products that bring in the most traffic, revenue or margin and add your own information: use cases, differences, experience and answers to customer questions. For the long tail, a well-structured specification table can already make a real difference.

Does the shop platform matter for SEO?

Less than people often think. What matters more is whether you control URLs, canonicals, filters, structured data and page speed. Check those points before choosing or switching platforms, because a platform change is always a migration with risks.

Does an online shop need a blog?

A blog is not a goal in itself, but buying guides and comparisons are valuable for shops. They answer research-phase questions, attract links more easily than product pages and let you pass authority to your categories through internal links.

Do I need an llms.txt file to appear in AI shopping?

No. Google states explicitly that Google Search ignores llms.txt and similar AI files. For visibility in AI answers, good product pages, an accurate Merchant Center feed and crawlable content carry far more weight.

Curious what SEO and GEO could do for your business?

Tell me where your business stands today and where you want to take it. In a short, no-obligation call we will work out which steps will make the biggest difference.

Get in touch →