Ecommerce SEO has a distribution problem. Effort concentrates on product pages because they are where the transaction happens, while category pages typically generate three to five times more organic revenue because they rank for high volume head terms and capture buyers before they have chosen a specific item.
A category page ranking for a broad term captures hundreds of related queries simultaneously. A product page ranks for that product. Both matter and the effort allocation across most stores is backwards.
Category Pages Need Content, Not Just a Grid
Most stores serve a category as a product grid with a generic H1 and nothing else. That page competes for a competitive head term with no unique content on it.
The placement debate around category copy has a reasonable resolution. Testing by large retailers consistently shows users want products visible immediately, which argues for a short introduction above the grid, somewhere in the region of fifty to a hundred words, and a longer editorial section below it.
That structure satisfies both constraints. Shoppers see products without scrolling past an essay, and the page carries substantive unique content for indexing.
What that editorial section should contain: how to choose within the category, what specifications actually matter for which use, and the distinctions buyers commonly confuse. That is genuinely useful and it is not manufacturer copy.
Product Descriptions Copied From the Manufacturer Are the Default Failure
Manufacturer supplied copy appears identically across every retailer stocking that item. Google filters duplicates and keeps whichever source it judges most authoritative, which is rarely a mid sized store.
Rewriting at scale is genuinely expensive and the prioritisation is straightforward: rewrite the products that carry the most revenue or the most search demand, and leave the long tail on manufacturer copy until capacity allows.
What a rewritten description should add: who the product suits, what problem it solves, what surprised actual customers, and how it differs from the obvious alternative. Those are things the manufacturer cannot write because they do not know your customers.
Faceted Navigation Is the Largest Technical Problem Most Stores Have
A catalogue with four facets of ten options each produces upwards of ten thousand possible URL combinations. Left unmanaged, that consumes crawl budget on near duplicate pages showing the same products in different orders.
JetOctopus data indicates the average ecommerce site has around 40 percent of its pages crawled less than once per month, which means a substantial share of inventory is effectively invisible for weeks at a time. Facet sprawl is a major contributor.
The handling that works, in tiers. Filter combinations with genuine independent search demand, like a colour and product type pairing people actually search, get treated as real indexable pages with their own titles. Medium value combinations canonicalise to the parent category, which consolidates signal while keeping the filter functional. Low value combinations, including sort order and view mode parameters, get noindex plus canonical.
Blocking facet parameters in robots.txt is the common instinct and it is usually wrong for the reason covered in the canonicalization guide: a blocked URL cannot be crawled, so its canonical is never read, so no consolidation happens.
Variants Need a Deliberate Decision
A product in eight colours and six sizes can generate dozens of URLs for what is functionally one item.
The default should be canonicalising variants to a single primary product URL, with selectors that do not change the URL. Split variants into separately indexable pages only where a specific variant has real independent search demand, which is a per product judgement rather than a blanket rule.
Indexing every variant by default is the more common error and it produces exactly the thin duplicate inventory that dilutes a domain.
Product Schema, Complete Rather Than Present
Product and Offer schema with price, availability, rating, and review count enables rich results and is close to mandatory for shopping surfaces.
The fields most often left empty are the ones that make a product identifiable across sources: GTIN, MPN, and brand. Without an identifier, your page describes a product that a system cannot confidently match to the same product elsewhere.
Schema must match the visible page. A markup price differing from the displayed price is the mismatch problem covered in the schema errors guide, and on ecommerce it happens routinely because feeds and page templates update on different schedules.
Discontinued Products Need a Policy, Not a Redirect Rule
Blanket redirecting every discontinued product to the homepage is treated as a soft failure by crawlers and it loses the specific answer someone was looking for.
Better handling: redirect to the closest genuine replacement where one exists. Where none exists, keep the page, mark availability as discontinued in both schema and visible content, and link to the category or nearest alternative. That answers the actual question, which is usually what replaced it.
Never leave a dead product page returning 200 with an out of stock message and in stock schema, which is the contradiction covered in the status codes guide.
Internal Linking Between Products and Categories
Ecommerce internal linking usually runs one direction, from category to product, through the grid. Products rarely link to each other or back to related categories in any structured way.
Related product modules, complementary item links, and links from product pages back to their category and to relevant buying guides all distribute signal and shorten crawl paths. The site architecture guide covers why click depth matters, and deep catalogues are where it bites hardest.
Support Content Matters More Than Most Stores Realise
Independent analysis by Aleyda Solis across several ecommerce subverticals found that support articles, size guides, policies, and educational content accounted for a meaningful share of pages cited in AI answers, reported in the range of 20 to 40 percent depending on vertical.
That finding has a classic SEO analogue. Buying guides, sizing information, care instructions, and comparison content capture research stage queries that product pages cannot rank for, and they feed internal links into the commercial pages.
Most stores treat this content as a support cost rather than an acquisition channel, which leaves the research stage of the buying journey uncontested.
Review Content Is the Cheapest Unique Content You Will Ever Get
Customer reviews solve the duplicate content problem structurally, because no other retailer has your reviews. A product page carrying forty genuine reviews has substantial unique content that nobody wrote.
Review markup earns star ratings in results, which affects click through rate independently of position. The review schema guide covers implementation, and the requirement is that aggregate ratings reflect the genuine corpus rather than a filtered selection.
Prompting for specifics rather than open feedback produces better content. Asking what size someone ordered and how it fitted produces text that answers a real query. Asking for a rating produces a number.
Product Images Carry More Weight Than Most Stores Give Them
Descriptive alt text, meaningful filenames, and explicit dimensions are standard practice and routinely skipped on catalogues where images are bulk imported from a supplier feed.
Image file size is also where ecommerce Core Web Vitals fail. Product pages carry more images than any other page type, and unoptimised images are the most common cause of Largest Contentful Paint failures at scale.
Where to Spend First
Schema completeness and unique copy on your highest revenue products, because both are low complexity and affect multiple surfaces.
Category page content, because the revenue multiple relative to product pages justifies it and most stores have none.
Facet handling, because unmanaged crawl waste caps everything else.
Core Web Vitals and heavy engineering work after those, since product pages carry heavy image loads and the fixes are real but expensive relative to the first three.
Monitoring Crawl and Index Health
Search Console's index coverage report is the fastest signal that facets have escaped. Indexed page counts substantially exceeding your actual product and category count means filter URLs are being indexed.
Log file analysis shows what is actually being crawled and how often, which index reports do not, and the log file guide covers the technique.
The NotionCue AI Crawler Audit checks what crawlers receive from specific URLs, which catches product pages where specifications render client side and are invisible in the server response.
Start your free NotionCue trial and audit a product page and a category page. Client side specification tables are one of the most common findings on ecommerce templates.
Fast diagnostic: compare your indexed page count in Search Console against your actual product plus category count. A large gap in either direction is a finding. Too many means facet sprawl. Too few means crawl or indexation problems on inventory that should rank.
Common Questions
Should out of stock products be removed?
Not if the product is returning. Keep the page, mark availability accurately in schema, and offer alternatives. Removing pages for temporary stock issues discards accumulated ranking signal you will want back.
How much unique product copy is enough?
Enough to describe the product specifically rather than generically. A hundred words of genuine specifics beats four hundred words of padded manufacturer copy, and the padding is what triggers duplicate filtering rather than the length.
Do category pages need to be long?
No. They need unique content that helps someone choose. Two to three hundred words of genuine buying guidance below the grid does more than a thousand words of keyword padded description.