How many articles does a B2B site need
Guide · SEO / AEO / GEO · 4 min read · last verified 2026-07-27
The number of articles a B2B site needs is the number of real questions its buyers ask, answered once each, answered well, and kept current. That is a size, not a target: it is discovered by mapping the questions, not chosen by picking a figure, and it comes out different for every market. Everything useful in content planning follows from taking that first sentence literally.
Count is a proxy, and a bad one
Ask how many articles a site needs and you are really asking something else: how much of what buyers want to know does this site answer? Article count is a proxy for that coverage, and like most proxies it detaches from the thing it stands for the moment it becomes the goal. A site can hold a large library and answer almost nothing, because the library grew by publishing whatever was easy to write. A site can be small and nearly complete, because its market's question set is genuinely small.
If one line from this piece survives, let it be this: a content library is sized by its market's question set, not by ambition. The productive planning question is never how many articles to ship this quarter but which asked-and-unanswered question each draft retires. If nobody can name the question a page answers, the page is volume, not coverage.
The question set is the unit of planning
Buyers in any market ask a finite, discoverable set of questions on the way to a decision: what the category is, how it differs from the categories next to it, what adopting it really costs in effort and risk, how it fails, how to choose between approaches, what the first months look like, how it fits what they already run. The set is not infinite, because buying processes are not infinite. It grows at the edges as the market moves, but the core is stable enough to map.
Mapping it is ordinary work rather than a mystery. Mine sales calls and support threads for the questions hiding behind them. Collect what real buyer questions actually sound like, in the buyer's words rather than your category's. Watch what AI assistants are asked in your market and what they answer, which is the raw material of a visibility-driven content plan. Some markets exhaust their core set quickly; complex multi-stakeholder markets sustain a far larger one, because every committee role asks its own version of every question. That is why the honest answer to the title question is different per market — and why a universal number offered by anyone with a content product to sell, Magrios included, should be read as marketing rather than planning.
Publishing past coverage does damage
The intuition that more publishing always helps does not survive contact with how selection works. When several of your own pages answer the same question with slight variations, anything choosing a source — a search index, an answer engine, a human skimming your blog — faces a set of weak candidates from your domain instead of one strong one. Your pages compete with each other, and the strongest possible page never gets written because its substance lies scattered across five near-duplicates. In observed answer-engine behaviour, thin pages that partially answer a question appear to be passed over in favour of pages that answer it completely in one place, which is exactly the page a fragmented library fails to contain.
Thin volume carries a second cost that holds regardless of any ranking mechanics: attention. Every published page is a standing maintenance obligation — its claims must stay true, its examples current, its links alive. A library built past coverage converts editorial attention into the upkeep of pages that never earned their place, and staleness spreads from exactly there.
After coverage, the verbs change
Reaching coverage does not end content work; it changes the verb. Three verbs replace publish. Refresh: the market moved and a good answer drifted out of date, and updating it preserves the standing the page has accumulated — which is why refreshing old content without losing AI citations is a craft of its own rather than a chore. Merge: the near-duplicates described above become the single strong page that should have existed, and the weak ones are retired to point at it. Deepen: a page that answers its question at the surface gains the missing layer — the worked example, the failure mode, the trade-off — that makes it the reference answer rather than an adequate one.
A mature library measures activity accordingly: the interesting number stops being pages added and becomes questions answered well. A single research scan can still fill a week of content work — but the week fills with refreshes and merges as often as with net-new pages, and that ratio shifting is a sign of health, not of slowing down.
How to know where you stand
Three observable signals locate a library on the path between too little and too much. The unanswered-question list: if mining calls, tickets, and assistant answers keeps surfacing real questions with no page, keep writing — coverage is still ahead of you. The cannibalization check: if searching your own site for a buyer question returns several partial answers, stop adding and start merging — the library is competing with itself. The decay test: if your most important pages contain claims nobody on the team would defend today, net-new publishing is being funded by decay, and the budget should move.
None of this produces a number to put on a slide, which is the point. A content library is finished the way a dictionary is finished: complete for now, and permanently under revision. Teams that internalize this stop asking how many articles they need and start asking which real questions they have not yet answered — a shorter list, and the only one worth working.