One page or two? How we decide from the search results

Every keyword tool produces a long list of related phrases, and it is tempting to give each one its own page. Do that and you get dozens of thin, near-identical pages competing with each other. Merge everything instead and you bury real differences between what people need. The question underneath both mistakes is the same: do these searches belong on one page, two pages, or an article?

Keyword similarity cannot answer that. Two phrases can look almost identical and want completely different things, or look different and want the same thing. What answers it is the search results themselves, because they show what Google has learned people want when they type each phrase.

The method

1. Demand. Google’s Keyword Planner shows how often each phrase is searched, and suggests related ones.
2. Questions. The People Also Ask box shows the questions buyers have around each phrase.
3. Compare the results. For each pair of candidate phrases, compare the first page of results: how many of the same sites appear, and just as important, what type of page ranks (service businesses, blog articles, software tools, courses, directories).
4. Check the AI answer. Where Google shows an AI Overview, compare which sources it cites.
5. Decide. Same intent, same audience, same kind of results: one page. Same topic but a different question: a section or a supporting article. Different intent, audience or location: a separate page.

The final call comes from putting these together, not from any single signal. Here is what that looked like on real searches, all checked in Australia in September 2026.

Low overlap, still one page

“Marketing strategy agency” and “marketing strategy consultant” shared only about a tenth of their page-one results. On overlap alone, that looks like two pages. But both result pages were the same type: real agency and consultancy sites, plus a business directory. The difference was agency (a team) against consultant (one person), not a difference in what the buyer wanted. One page.

The AI answers disagreed, which was worth noting. The agency search’s AI Overview cited videos, a university and a government site. The consultant search’s cited consultancies directly. We kept the one-page decision, because what actually ranks, and the fact that this consultancy is one person, outweighed one oddly generic AI answer. When signals disagree, say so and weigh them, rather than picking whichever gives the cleaner answer.

Nothing in common: separate content

“Content marketing strategy” and “content marketing examples” shared no meaningful results at all, and the types differed completely. The first returned universities, an academic journal and videos. The second returned marketing blogs and list articles. “Examples” is article territory, not part of a service page.

Big volume, wrong intent

“Customer journey mapping” is searched around 1,900 times a month in Australia, about seventeen times as often as “marketing strategy agency”. It looked like an obvious page to build. The results said otherwise: UX software tools, a research firm and a business school’s online courses, with every People Also Ask question shaped like a how-to guide and no AI Overview at all. People searching it want a method or a tool, not a consultant. It is a good article topic and a poor service page.

The phrase that sounded right and wasn’t

“Automation consultant” sounds like exactly what a business automation service should target. Google reads “consultant” as the important word and returns lactation consultants, workplace safety bodies and heritage consultants. No amount of good content makes that a useful target. Always look at the results before building around a phrase.

Check the data, too

Search data can be wrong without looking wrong. In one batch of 23 searches, a data provider returned results for “sales automation” that were a pesticide regulator and the tax office, with no error attached. Three other searches in the same batch came back similarly broken. The same searches through a second provider looked normal. If a result page does not make sense for the query, cross-check it before drawing a conclusion from it.

What it changed for this site

We ran every page planned for this website through the same process. Twenty-six planned pages became sixteen: ten merged into a sibling page or were dropped, and one that had only been under consideration was promoted to a full page. Several of the dropped pages became article topics instead.

The short version: keyword lists tell you what people type. The search results tell you what they want. Build pages around the second.

If you would like the same process applied to your own site, it is how every content plan we write starts.