Primary Sources Get Cited, Summaries Get Skimmed

When a writer needs to support a claim, they try to cite where the claim came from. That instinct is nearly universal and it decides who gets the link: if your page is a summary of somebody else’s work, the honest writer reads you, learns the fact, follows your source, and cites the source instead.

Which means summarising is structurally unciteable. Not badly written — unciteable by design. You can produce the clearest, best-organised restatement of a finding on the internet and still collect nothing, because the fact isn’t yours to be credited for.

The citation chain

Watch how a fact travels. Somebody measures something and publishes it. A trade publication writes it up. A dozen blogs write up the write-up. Then hundreds of pages mention it in passing, each one linking whichever version they happened to read.

Over time, the links concentrate at the origin. Every writer who checks — and a meaningful share do check, because misattributing a fact is embarrassing — walks one or two hops upstream and cites there. The intermediate summaries are read constantly and cited rarely. They’re plumbing. Traffic flows through them; credit doesn’t stop there.

There’s a real exception worth naming. Sometimes the citations pile up on whichever summary ranks best, because the writer searched the fact, landed on the highest result, and cited it without looking further. That happens, it happens a lot, and it is a fragile position to build on. It evaporates the moment somebody looks properly, and it puts you in the awkward spot of being credited for a fact you didn’t produce — which is exactly the credit a careful writer will take away from you and hand upstream. Ranking for someone else’s finding is a rented position. Owning the finding is not.

What counts as being primary

Being primary is narrower than “having something original to say,” and broader than “running a research study.” You are the origin of a fact when:

You measured it. You counted, timed, surveyed, sampled or aggregated something, and the number did not exist before you produced it.
You built it. A tool’s output, a dataset you assembled from scattered public records, a comparison nobody had compiled — the assembly work is itself origination.
You were present. You attended the thing, ran the experiment, did the migration, watched the failure. First-hand accounts are primary even when they’re informal.
You are the authority who decides it. Your own product’s behaviour, your own pricing, your own policy, your own limits and terms. Nobody can be a better source about your system than you are, and this is the cheapest primary status available — it requires documenting what you already know.
You stated a distinction first, and clearly. Naming a difference that people previously conflated is origination. When later writers need to explain that difference, they cite whoever drew the line.

Notice how much of that list is documentation rather than research. Most sites underrate the last two entries badly.

The “one step upstream” exercise

Take your five best pages. For each one, ask a blunt question: what is this downstream of?

If the answer is “an industry report,” “a vendor’s documentation,” “a study from four years ago,” or “the other articles about this topic,” then you are an intermediate node on somebody else’s citation chain. That’s the diagnosis. The follow-up question is the useful one: is there something upstream of the same topic that you could own instead?

Often there is, and it’s smaller than the page you have. Instead of summarising what a standard says, publish what happened when you implemented it. Instead of restating what a report found, publish the equivalent number from your own operation. Instead of explaining a concept everyone explains, publish the edge case you hit that nobody documents. The upstream version is usually narrower, less impressive-looking, and vastly more citable.

Where summarising is legitimate

Don’t take this as an argument against ever explaining anything. Summaries do real work, and some of the most valuable pages on a site are summaries by design:

Explainers for your own readers, who came to you for an orientation and would be poorly served by being sent to a dense primary source.
Onboarding and internal training material, where the point is getting someone up to speed, not being cited.
Landing pages and overviews that organise a topic so a reader can find the right sub-page.
Genuine synthesis, where you’ve read twenty sources and the comparison is new even though no individual fact is. This shades into being primary, and the more the structure is yours, the more citable it gets.

The mistake isn’t writing summaries. It’s writing summaries and then being surprised they don’t earn links, or worse, building an entire content plan out of them and calling it a link-building strategy. Judge a summary by whether it serves your readers. Judge a link asset by whether the fact in it starts with you.

The failure mode

The clearest version of this failure is the industry statistics page: gather the well-known figures everyone in your field quotes, rewrite them into a list, and publish. It looks like a link asset. It is often pitched as one.

What it actually produces is a page nearly identical to several dozen others, containing no fact that originates with you, competing on ranking alone for the right to be miscited. When a writer checks any single line on it, the link goes to the original source — which is the correct outcome, and it was always going to be the outcome. Meanwhile the page ages badly, because you have no way to update figures you didn’t produce, and no way to know when they stop being true. The problems with statistics roundups go further than duplication, but duplication alone is enough to sink them.

What to do next

Stop asking what you could explain, and start asking what you could be the origin of. That’s a shorter list and a more useful one.

The cheapest primary source almost every site already owns is its own operational knowledge: what your system does, what your customers ask, what your process actually produces, measured across everyone you serve. Nobody else can publish those numbers, which makes them the strongest starting point you have. That’s the subject of the original data you already have.

And before you build anything, check the candidate against the four conditions that decide whether a page earns links — being primary satisfies one of them well, but only one.