Deciding Whether to Publish a Page

Most website pages that underperform were not built badly. They were built without a clear answer to a more basic question — should this page exist as a separate page at all, on this site, right now? This guide walks through the qualifying decision before any time is spent on drafting, designing, or briefing an AI agent.

Key Takeaways:

• A page idea is an opportunity to qualify, not an instruction to publish.
• The first filter is the reader task — name the specific job a real reader brings to the page.
• The second filter is overlap — search the site honestly for pages that already cover this or a near-duplicate angle.
• The third filter is distinct value — what does this page give that a generic summary cannot.
• The fourth filter is silo fit — the page either extends an existing silo or it does not get built yet.
• A failed qualification is a saved week of work, not a problem.

The cost of skipping the decide phase

Every page on a site competes for crawl budget, internal-link equity, the reader’s attention, and the site’s topical clarity. Pages that exist without a clear reader task tend to dilute related pages that do. The practitioner observation — not a Google-confirmed mechanic — is that a site with eighty strong articles and twenty hollow ones often appears to perform worse over time than a site with eighty strong articles and no hollow ones, because the hollow pages absorb inbound links and internal links that the stronger work would otherwise receive.

This is the practical case for treating publication as a decision rather than a default. The cheapest moment to discard a weak page is before it exists.

Filter one — the reader task

A page has a real reader task when you can finish the sentence “someone arrives at this page because they are trying to ___” with a concrete verb and a concrete object. “Decide between two managed hosting plans for a small WordPress site” is a reader task. “Learn about hosting” is not.

When the task is hard to name in one sentence, the article almost always splits later into two narrower pages, or merges back into a page that already exists. Doing that work in the decide phase is faster than doing it in the diagnose phase six months after publication.

It is also acceptable for the answer to be “no one specific is trying to do anything specific on this page.” That answer disqualifies the idea and saves the time.

Filter two — overlap with what the site already covers

Before drafting, search the site’s existing content for the topic and for adjacent angles, not just the proposed title. Site-search inside Google (site:yourdomain.com topic) is more honest than the in-CMS search because it shows what is actually indexed. Read the top three matches end to end. If one of them already addresses the reader task you named in filter one, the right output is usually an expansion or a refresh of that existing page, not a new URL.

New URLs introduced on top of overlapping content tend to cannibalise — both pages ranking in mediocre positions for the same query, neither earning a click that a single stronger page would have earned alone.

If the overlap is partial — the existing page covers the topic for a different audience, or for a different stack, or at a different stage of the lifecycle — that is a case for a new page, with explicit cross-links in both directions to make the distinction obvious to readers and to search engines.

Filter three — distinct value beyond a generic summary

Imagine the reader has already asked an AI assistant the same question and read its answer. What does your page give them that the assistant could not? A useful answer involves at least one of these — a perspective grounded in lived experience, a worked example with a specific tool or configuration, a comparison drawn from running both options, a decision framework that holds up across scenarios the assistant flattens, an opinion you are willing to defend, or evidence the assistant does not have access to.

A page that cannot articulate a distinct-value case before it is written is not necessarily a bad page, but it is a page at risk of becoming generic content. The decide phase is when that risk is cheapest to recognise.

Filter four — silo fit

A site organised into silos earns topical clarity from that organisation. A page that does not belong inside any existing silo creates a structural question — does the page belong on a different site, does it belong inside a new silo that needs at least a handful of supporting pages before it is worth opening, or is the proposed page actually the seed of a silo that does not yet exist?

The honest answer is sometimes “this idea is real but premature.” A single page on a topic that has no surrounding silo on the site is harder to connect internally than the same page would be inside an existing topical cluster, and tends to perform less well as a result. Holding the idea until a small cluster of related pages can be planned together is often the stronger move.

When all four filters pass

A page idea that passes all four filters is ready to move into the build phase. The output of the decide phase is not a draft — it is a short statement that records the reader task, the overlap analysis, the distinct-value case, and the silo location. That statement is what the build phase reads as its brief, and what the evaluate phase later checks the published page against.

If the page is being produced with help from an AI agent, the decide-phase statement is also what the agent is given to constrain its work. An agent that begins drafting before this statement exists is doing the wrong work fast.

When a filter fails

A failed filter does not necessarily kill the idea. Filter one might fail because the original topic was too broad, in which case narrowing the reader task often unblocks it. Filter two might fail because an existing page does the job — in which case the action is to improve that page, not write a new one. Filter three might fail because the distinct value has not been thought through yet — and a different angle on the same topic might pass. Filter four might fail because the silo does not exist yet — and the idea might be worth holding until it has company.

What does not happen is publishing anyway. The qualifying decision is the same whether the page is written by a person, by an agent, or by both.