Content that earns links in ecommerce

A product page is a commercial document. It exists to convert one specific visitor, and it is written in the second person by an interested party. No journalist, blogger, forum moderator or Wikipedia editor is going to cite it as a source, and the ones who do are being paid, which is a different problem addressed further down.

This is not a failure of quality. A beautifully written collection page for men's waterproof jackets is doing its job perfectly and will accumulate no links at all, forever. The mistake is concluding that the page needs improving. It needs internal links from things that do earn attention, and it needs to rank on relevance and merchandising strength — which is a separate discipline with a separate playbook.

So stop measuring commercial pages by referring domains. The link work happens elsewhere on the site, and its job is to raise the whole domain and then pass equity inward deliberately.

Before anything is written, answer two questions in one sentence: which article cites this, and who is writing it?

"A guide to choosing a mattress" fails. Nobody writing anything needs to link to it, because a hundred equivalents exist and none of them is the authoritative version of anything.

"The average return rate by size across 40,000 mattress orders" passes, if you actually have that data. A journalist writing about fit, a trade publication writing about returns, a competitor's marketing team writing about sizing — all of them need a number, and you are the only person who has one.

The test works because it converts a vague content brief into a distribution plan. If you can name the article, you can name the publication, which means you know who to send it to and what subject line to use. If you cannot name the article, no amount of production budget will produce a link, and the piece is being commissioned on hope.

Each of these is buildable by a mid-market merchant without an agency retainer. They are listed in rough order of how underused they are.

Data only you have. Your order book, returns, search logs and support tickets contain facts nobody else can produce. Sizing patterns by region, seasonality by category, the questions customers ask before buying. The cost is analytical rather than editorial, and the two rules are: aggregate properly so nothing identifies a customer, and publish the method alongside the number so it is quotable. Do not publish a figure you cannot show the working for — an unsourced statistic gets quoted once and then contradicted, and you become the source people stop trusting.

A reference somebody needs repeatedly. A compatibility matrix, a materials comparison, a care standard, a conversion table for the sizes your industry refuses to standardise. These are dull and they accumulate links for years, because every new article on the topic needs somewhere to point. They also age well, which nothing else here does.

A tool that answers a question faster than reading. A calculator, a fit finder, a coverage estimator. Genuinely effective and genuinely expensive — not to build, but to keep working. A broken tool with three hundred links is worse than no tool, so only build one if there is a named owner for it in two years' time.

Specifics about how things are made. Named factories, actual materials, the reason a component costs what it does, what went wrong in a production run. This is the cheapest of the four because you already know it, and the hardest because it requires saying something a competitor could use. That is precisely why it gets cited.

Worth reading the actual policy rather than the received wisdom, because the boundary is clearer than most agencies present it.

Google defines link spam as "the practice of creating links to or from a site primarily for the purpose of manipulating search rankings", and the listed examples are specific: "exchanging money for links, or posts that contain links", "exchanging goods or services for links", "excessive link exchanges" and "advertorials or native advertising where payment is received for articles that include links that pass ranking credit".

The exception matters as much as the rule. Google acknowledges that "buying and selling links is a normal part of the economy of the web for advertising and sponsorship purposes" — provided the links carry rel="nofollow" or rel="sponsored". Paying for a placement is a marketing decision you are allowed to make. Paying for a placement that passes ranking credit is the violation. Anyone selling you the second while describing it as the first is selling you the risk as well.

Two newer policies are worth knowing because they retire tactics that were fashionable recently. Site reputation abuse covers third-party content "published on a host site mainly because of that host's already-established ranking signals" — the subfolder rented on a newspaper's domain. Scaled content abuse covers "many pages generated for the primary purpose of manipulating search rankings and not helping users", explicitly including "using generative AI tools or other similar tools to generate many pages without adding value for users". Between them they describe most of what is currently pitched as a scalable content programme.

Most merchants do not have an ideas problem. They have a publishing problem, and it is structural.

A data piece wants charts and a methodology block. A reference wants a sortable table. A transparency piece wants a full-bleed photograph and a pull quote. None of those exists in the theme, so each one becomes a development ticket, and a development ticket means the piece ships in six weeks or does not ship at all. The marketing team learns this after two attempts and quietly stops proposing anything that is not a blog post, which is why so much ecommerce content is a blog post about a topic that deserved a table.

The fix is a section library: a set of composable blocks — comparison table, data panel, stepped process, figure with caption, quote — that a marketer arranges without a developer. It changes what can be commissioned. When a format costs an afternoon instead of a sprint, the ideas that were never worth the ticket become worth trying, and link-earning content is disproportionately made of exactly those formats.

This is also the honest reason to fix publishing speed before commissioning a campaign. An agency retainer producing ideas your site cannot publish is money spent on frustration.

Count referring domains to the specific asset, and count them at 30, 90 and 365 days. That is the whole measurement. Domain-authority scores are vendor models, useful as a rough sort and worthless as a target, and traffic to a link asset is mostly irrelevant — a reference table with 200 monthly visits and 40 referring domains is doing its job.

Then be honest about the distribution of outcomes. Most pieces earn nothing. This is normal and it is why link-earning content has to be budgeted as a portfolio rather than as a series of individually justified projects: commission six, expect one to work, and let that one carry the programme. A quarterly review that kills the effort after two failures has misunderstood the shape of the thing.

You will find statistics quoted about what proportion of content earns no links at all. We are not repeating any of them here, because the ones in circulation come from crawls of different corpora with different definitions and we have not verified any against its source. The directional claim — that most published content is never linked to — is uncontroversial enough without a decimal point attached.

The annual industry survey. Everybody has one, they all show the same thing, and the response rate required to make the numbers meaningful is larger than most merchants can reach. Your order data is a better dataset than a survey you have to recruit for, and it is already collected.

The paid listicle placement. "Best 10 X" articles are sold, the link is what you are paying for, and the policy above says exactly what that is. If you want the traffic, buy it as advertising with the link qualified — that is permitted and it is measurable.

And the content calendar of forty pieces a quarter. Volume was never the mechanism here. One reference table your industry ends up using is worth more than a year of posts, and it does not need refreshing every six months to stay alive.

• Shopify section library
• Internal linking, done deliberately
• Collection pages are your best SEO asset