SEO for Ecommerce Stores: You Are Selling the Same Object as Everyone Else

Two other pages here already handle most of what an online store needs, and this one assumes you can reach them by link instead of reading it twice. Platform tiers, the real cost of catalog work, replatforming, and what it takes to be legible to AI shopping assistants all belong to our ecommerce website cost guide, which is why no figure appears below. Merchandising automation, abandoned-cart recovery and product-feed operations sit on the AI for ecommerce businesses page. This piece is only about search.

It is also the one guide on this site where the usual advice does not transfer. Every other SEO piece we have published rests on a geographic moat — a map result, a service area, a proximity advantage that puts a ceiling on how many businesses can realistically compete for a query. An online store has none of that. There is no radius protecting you, and the competitor set for any given product is national and frequently includes the company that made it.

The Manufacturer Wrote Your Product Page, and Everyone Else’s

Start with the uncomfortable structural fact. If you resell rather than manufacture, your product page most likely carries the supplier’s description, the supplier’s photography and the supplier’s specification table. So does the page on the brand’s own site, the listing on two marketplaces, and the identical page at every other authorised dealer. Search engines are looking at several dozen documents that say the same thing and choosing between them on signals that have nothing to do with the text — and a small store rarely wins that comparison.

Price is usually unavailable as a differentiator too. In a great many categories the brand enforces a minimum advertised price, which means the number on your page and the number on your competitor’s page are the same number by contract. Strip out text you did not write and a price you cannot change and there is very little of your business left on the page. That is the actual problem, and it is not solved by adding more product pages.

What “your own content” means in practice

Not a rewritten paragraph of marketing copy. The things that are genuinely yours are the ones that come from selling the product rather than from listing it: which of the three versions people end up returning, what it does not fit, the question your support inbox gets about it every week, how it compares to the other two you stock, what a customer should buy instead if they are using it for something it was not designed for. A supplier cannot write any of that because they do not know it, and the marketplace cannot because it does not have a point of view. This is also the distinction our guide to being recommended by AI assistants turns on — a system summarising a category has no use for the fortieth copy of the manufacturer’s paragraph.

The Category Page Is a Page About a Decision

Category pages are where non-branded ecommerce search actually lands, and most of them are a heading, a grid, and a block of text stuffed underneath the products where nobody reads it. That block exists because someone was told the page needed words on it, which is a different goal from the page being useful.

The shopper arriving on a category page from search has not chosen a product. Frequently they cannot yet, because they do not know which attribute matters. A grid of twenty-six options is useless to that person and mildly punishing — it presents the decision as already made and asks them to scroll.

What belongs above the grid

The one or two attributes that actually determine the choice in this category, stated plainly, with the trade-off named. Which subgroup suits which use. The mistake people make buying this, which you know because you process the returns. A short route into the grid for each of the two or three common cases. Then the products. This is a few hundred words of genuine expertise rather than a thousand words of filler, and it is the only version of a category page that competes with a marketplace, because a marketplace cannot make a recommendation without undermining the sellers it hosts.

The filter trap underneath it

Faceted navigation is the most common technical disaster in ecommerce search, and it is invisible from the storefront. Five filter groups combine into an enormous number of distinct URLs, most of which return a handful of products and a page that is nearly identical to a hundred of its neighbours. Left open, a crawler spends its attention on that combinatorial sprawl instead of on the pages you care about, and a store with a few thousand products can present a few hundred thousand crawlable addresses.

The rule that survives contact with reality: a filtered view deserves a real indexable page only when people search for that exact combination and the result is a coherent set worth landing on. A size, a fitment, a material, a use case — those can be genuine pages with their own introduction. A colour crossed with a price band crossed with a rating is not a search anyone performs, and it should be usable by shoppers and closed to crawlers. How you close it differs by platform, which is exactly why this is worth confirming on yours rather than inheriting a rule written for somebody else’s.

Compatibility Is the Question Marketplaces Answer Worst

In any category where parts have to fit something — vehicles, machinery, appliances, instruments, tanks, printers, cameras, bicycles — a large share of search is a fitment question, and it is the highest-intent question in the whole category. Somebody asking whether a part fits a specific model is not browsing. They are one confident answer away from buying, and they are frequently mid-repair.

Marketplaces handle this badly and structurally so. Their fitment data is supplied by thousands of sellers at uneven quality, it is presented as a lookup widget rather than as an explanation, and the answer is almost never accompanied by the qualification that makes it true — fits this model but only before the mid-cycle revision, fits but needs a different bracket, fits mechanically and voids something if you do it. That qualification is precisely what a specialist retailer knows and precisely what the anxious buyer is searching for. Publishing it as real pages rather than as a filter is one of the few places a small store can beat a marketplace outright on a commercial query.

The same logic runs through sizing in apparel and equipment, dosage or coverage in consumables, and capacity in anything that holds something else. Any question whose wrong answer causes a return is worth a page, because you are already paying for the wrong answer at the other end.

Your Best Pages Are Attached to Inventory That Expires

This is the problem no other business on this site has. A law firm’s practice page does not go out of stock. A store’s ranked URL is tied to a product that will eventually be seasonal, unavailable, replaced or discontinued, and what happens at that moment decides whether years of accumulated search equity survive.

The default behaviour in most stores is deletion. The item leaves the catalog, the page returns a not-found, the rankings and any links pointing at it evaporate, and nobody connects the traffic decline three months later to an inventory decision made by someone in a different department. There are three distinct situations here and they have three different correct answers.

Temporarily unavailable

Keep the page. Say plainly that it is out of stock, give a restock expectation if you have an honest one, offer to notify, and show the nearest alternatives. A page that tells the truth about availability is more useful than one that vanishes, and it is still there when the stock is. Removing a page for a six-week gap in supply trades a durable asset for a tidier catalog.

Discontinued with a successor

Redirect to the successor, and say on the destination that it replaces the old model and what changed between them. The person searching the old model number almost always wants the new one and does not know it exists yet. That sentence — what replaced what, and what is different — is a page in its own right in categories with model cycles, and it is one manufacturers write badly when they write it at all.

Discontinued with nothing equivalent

The hardest case and the one most often handled worst. If the URL has links or steady demand, a page explaining that the item is no longer made, why, and what the closest current options are will usually serve the visitor and the business better than a not-found. Redirecting it to a broadly related category instead is the tempting shortcut and it is worse than either option, because the visitor arrives somewhere that does not acknowledge what they asked for.

Reviews and Questions Are the Only Text on the Page That Is Yours

On a page otherwise assembled from supplier material, the customer content is the sole original component. It is also the only source of the vocabulary your buyers actually use, which is reliably not the vocabulary your category manager uses. People describe problems, uses, fit, noise, durability and disappointment in words that map onto real searches, and those words arrive on your site at no editorial cost.

Answer the questions publicly and in full. A support reply sent privately helps one person; the same reply published under the product helps every subsequent visitor asking the same thing and gives the page substance no competitor selling the identical item possesses. Two constraints, both of which belong to your own counsel rather than to us: fabricated reviews and undisclosed incentivised ones are a genuine legal exposure in the United States and not merely a reputational risk, and review volume does not rescue a page that has nothing else on it.

Stop Trying to Beat the Marketplace at Its Own Search

A specific branded product query — the exact model number, the exact name — is a transactional search that the large marketplaces are built to win, and they carry purchase and behaviour signals that no independent store will accumulate for that query. Competing there is not a strategy, it is a subscription to disappointment.

What sits upstream of that search is unclaimed, because it is expensive to produce and nobody is paid to produce it. Which one should I get. Does this fit mine. What is the difference between these two. What do I need for this job. Why did the last one fail. Those searches precede the branded query, they are answered badly across most categories, and the person who answers them is frequently the one who gets the order — not because they ranked for the model number, but because the shopper stopped needing to search for it. That is the same pattern we describe in why traffic does not become leads, arriving from the opposite direction: the visit is worth something only when the page did the work before the visitor decided.

The Order to Do This In

First, stop the bleeding: settle what happens to a URL when a product goes away, and close the faceted sprawl. Both are one-time decisions that stop ongoing losses, and neither produces anything visible, which is why they are perpetually deferred.

Second, write the pages only you can write — the comparison across the brands you stock, the fitment answers with their qualifications attached, the buying guide for the decision your customers get wrong. Start with the categories where you carry the deepest range, because that is where your knowledge is real and where the page will not read as guesswork.

Third, rebuild your highest-traffic category pages as pages about a decision. Fourth, publish the questions your support inbox already answers. And judge all of it by revenue arriving on non-branded landing pages rather than by rankings on product URLs you were never positioned to hold. We build this with Claude Code, which is how catalog-scale work — templates applied across thousands of pages, structured data emitted consistently, redirects mapped in bulk — gets done in days instead of being quoted as a quarter and then abandoned.