How Wide Should One Page Be
The comprehensive guide does not win. Neither does the single question page. What carries volume is one topic, covered deeply enough to answer the cluster of questions inside it.
Key Takeaways
- 84.3% of quoted pages win exactly one question. 93.9% sit inside a single subject area.
- But the 15.7% that win several questions produce 52.4% of all citations.
- A page that answers several related questions is quoted 21 times a month against 3.6 for a single question page.
- Two thirds of those multi question winners still stay inside one subject.
- Plan by question cluster. Not one page per question, and not one page per category.
Why we looked into this
Every content team eventually has to choose an architecture. One option is the comprehensive guide: a single large page that covers a whole subject and attracts everything related to it. The other is the narrow page: one page per specific question, each one short and precisely aimed.
The choice determines how a team spends its year. The guide model concentrates effort into a handful of major pages that get maintained and expanded. The narrow model produces many smaller pages and needs a machine to keep them organised. Getting it wrong is expensive in a way that only becomes visible twelve months later.
AI answers are assembled per question, which suggests narrow pages should win. But the same logic taken too far produces hundreds of thin pages, which is its own failure. We wanted to find where the line actually sits.
How we tested it
For each of the 9,702 quoted pages we counted two things. How many distinct questions earned it a citation, and how many distinct subject areas those questions belonged to.
The second measure is the more interesting one. A page quoted for six questions that all sit inside one subject is doing something very different from a page quoted for six questions spread across the whole category. The first is depth. The second is breadth. Counting questions alone cannot tell them apart, which is why most analysis of this stops at the wrong answer.
We then looked at where the citation volume actually accumulates, because the count of pages and the share of citations turn out to tell opposite stories.
What we found
Most pages win exactly one question
Of the 9,702 pages AI quoted, 84.3% were quoted for exactly one question. And 93.9% sit inside a single subject area. On the face of it, that is a decisive verdict for narrow pages and against the comprehensive guide.
But a small group carries half the volume
Then the picture turns. The 15.7% of pages that win more than one question produce 52.4% of all citations. About 1,500 pages earn slightly more than the other 8,200 put together.
The reason sits in that last comparison. A page that answers one question is quoted around 3.6 times over the month. A page that answers several is quoted around 21 times, almost six times more.
The effect compounds rather than adds. Answering a second and third related question does not simply bring a little more visibility, it makes the page eligible for several different answers instead of one, and the returns multiply.
The winners stay inside one subject
This is the part that resolves the apparent contradiction. Of the pages that win several questions, about two thirds stay within a single subject area. They are not sprawling across a category. They are covering one topic thoroughly enough that a cluster of nearby questions all land on the same page.
So neither extreme is right. A single narrow answer page usually earns its one citation and stops there. A comprehensive guide spanning a whole category does not win either. What carries volume is one topic, covered deeply enough to resolve the several questions that sit next to each other inside it.
Why this happens
An AI engine answers a question by finding a passage that resolves it. A page covering one topic in depth contains several such passages, each aimed at a different nearby question, and each fully supported by the surrounding material.
A page covering a whole category contains the same number of passages spread far thinner, and each one has to compete against a specialist page that treats that question properly. A page answering exactly one question has a single passage and no second chance.
The cluster page wins because it is genuinely the best answer to several questions at once, rather than a compromise across many.
What this means for your business
The planning unit should be the question cluster, not the keyword and not the category.
In practice that means starting from the questions your buyers actually ask, grouping the ones that a single well built page could genuinely resolve together, and building one page per cluster. Not one page per question, which produces thin pages and stops at 3.6 citations each. Not one page per category, which produces a guide that wins nothing in particular.
The economics are worth stating plainly. Ten pages each answering one narrow question earn roughly 3.6 citations each. A smaller number of cluster pages earn roughly 21 each. That is a materially better return on the same production effort, and it argues for fewer, better scoped pages rather than more of them.
There is also a maintenance argument. Fewer pages means each one can be kept genuinely current, and currency matters in a channel that re evaluates its sources continuously. A library of hundreds of thin pages ages badly and nobody has time to refresh it.
One caution on the other side. Two thirds of multi question winners stay inside a single subject, which is a strong tendency rather than a law. A third do span topics. Treat the single topic rule as the default and the exception as something you choose deliberately, not something that happens because a page grew without a plan.
Building that cluster map is where we start a content plan at Novastacks.
Methodology
89 questions agreed with chief executives and business owners as ones their companies need to monitor, put to ChatGPT, Google AI Overview and Perplexity every day through August 2026. 6,601 answers and 61,536 citations across 9,702 pages.
Subject area breadth is measured using the topic groupings that were assigned to each question when the study was designed, so the categories are independent of anything we observed afterwards. Some entries among the widest reaching pages are citations to a website rather than to a specific page, which inflates apparent breadth, and these are identified separately in our data. The questions sit within business software categories, so breadth ceilings here are naturally higher than they would be across an unrelated sample. All brands are anonymised.
Frequently Asked Questions
Should I write one big guide or many small pages?
Neither extreme. 84.3% of cited pages win exactly one question and earn about 3.6 citations. The pages carrying real volume answer a cluster of related questions inside a single topic and earn about 21.
How many questions should one page answer?
Enough that a cluster of closely related questions all land on it. Pages winning several questions produce 52.4% of all citations, and about two thirds of them stay within one subject area.
Are pillar pages dead for AI search?
A broad guide spanning a whole category does not win, but neither does a single narrow answer page. The shape that carries volume is one topic covered thoroughly enough to resolve several nearby questions.
How much more traffic does a cluster page earn?
In our data a page answering several related questions was quoted about 21 times over the month, against 3.6 for a page answering one. That is close to six times more.
How was this measured?
For each of 9,702 quoted pages we counted the distinct questions that earned it a citation and the distinct subject areas those questions belonged to, across 61,536 citations recorded daily through August 2026.
What Does Your Question Cluster Map Look Like?
We group the questions your buyers ask into the clusters a single page could genuinely own, then show you which clusters you cover and which you have left open.