The short answer
A query cluster is the set of different phrasings that all express one underlying need, and the current best practice is one strong page per cluster instead of one page per phrasing. The reason is that search engines and AI assistants match meaning, not exact strings.
Take three searches: "how long do roof shingles last," "roof shingle lifespan" and "when to replace asphalt shingles." Typed as keywords they look like three targets. As needs, they are one: the person wants to know how long shingles last and when to act. A single page that answers that well serves all three, and probably a dozen more variants you never thought of.
The older approach would make three pages, each tuned to its phrase. That tends to produce thin, repetitive pages that compete with each other and say less than one good page would. The current approach groups the phrasings, builds one complete answer and uses the variants naturally in headings and text.
This page explains the difference, when single-keyword pages still make sense and how to build a cluster from data you already have. It is closely related to how people now phrase things, which we cover in conversational queries versus traditional search queries.
The old way and the current way
The old way treated each keyword as a separate target with its own page, and the current way treats each need as a target served by one page. The change followed how search engines learned to read language.
Years ago, engines leaned heavily on matching the words in a query to the words on a page. If someone searched "cheap flights Boston," a page with that exact phrase had an advantage over one that said "low-cost airfare from Boston." So practitioners built a page for each phrasing, repeated the phrase and hoped. The tactic worked well enough to become a habit.
Search engines now model meaning. Google has said for years that its systems understand that different words can express the same intent. AI assistants go further, since they are built on language models that treat "shingle lifespan" and "how long roofs last" as the same question. When an engine matches by meaning, near-duplicate pages split your signals and add nothing. A reader landing on the third near-copy gets the same answer in slightly different words.
The diagram shows the contrast. One page per keyword spreads effort across many thin pages. One page per cluster concentrates it. This does not mean you stop caring what people type. It means you use what they type to understand needs, then serve each need once, properly.
The related idea of grouping content into hubs and spokes is covered in our glossary entry on topic clusters, which works at a larger scale than the query clusters discussed here.
What counts as one cluster
Phrasings belong in the same cluster when the same page would fully satisfy the person behind each one. That is the test, and it is more reliable than comparing words.
Ask: if a person typed this phrasing and landed on that page, would they get what they wanted? "Roof shingle lifespan" and "how long do shingles last" pass. "Roof shingle cost" does not, even though it shares words, because the person wants a price, not a lifespan. That is a different cluster and deserves its own page.
The strongest tell is shared search intent. Our glossary entry on search intent separates the main kinds: finding information, comparing options, taking a specific action, and buying. Phrasings with the same intent and the same subject cluster together. Phrasings with a different intent split off, even when they sound close.
Another check: look at the pages Google shows for each phrasing. If the top results are largely the same pages for two phrasings, Google treats them as one need, and you can too. If the top results differ a lot, they are separate needs, and one page will not win both.
A good cluster page also answers the neighbors. Someone who asked how long shingles last often wants to know what shortens it and how to tell when it is time. Those follow-ups belong on the same page, under their own headings.
When a single-keyword page still makes sense
Give a phrasing its own page when it reflects a distinct intent that one page cannot satisfy together with the rest. Clustering is a rule about shared meaning, not a ban on separate pages.
Distinct intent. "Roof repair" and "roof replacement" overlap in vocabulary but differ in the decision. A person comparing a repair to a full replacement needs a page that helps them choose. A person ready to book a replacement wants a service page. Forcing both onto one page makes it worse for each.
Distinct audience. "Gutter cleaning for homeowners" and "gutter maintenance contracts for property managers" need different language, proof and next steps, even if the subject is the same.
Distinct location or product. A service offered in three cities, or a product in three real variants with different specs, can warrant separate pages, as long as each has its own real content. A set of city pages with only the name swapped is the thin-duplicate problem again.
Brand and navigational terms. A page for your own brand name or a specific product model is its own target.
The question to ask is whether a reader of the merged page would feel something was missing. If yes, split it. If the pages would say nearly the same thing, merge them. Related guidance on avoiding near-duplicates appears in duplicate content.
How to build a cluster from Search Console
Your own Google Search Console data lists the real phrasings people used to reach you, which makes it the best raw material for clusters. It is free and already collecting.
Open the Performance report and export the queries for the last sixteen months, which is the span Search Console keeps. Put them in a spreadsheet with the page each one landed on, impressions and clicks. If you have never used it, our glossary entry on Google Search Console explains the basics.
Step one, sort by subject. Group queries that clearly concern the same subject, for example everything about shingle lifespan, shingle cost and shingle brands.
Step two, split by intent within each subject. Inside the shingle group, separate "how long do they last" questions from "how much do they cost" questions from "which brand is best."
Step three, name each cluster after its need, not after a keyword. "How long shingles last and when to replace them" is a cluster name. "Shingle lifespan" is a keyword.
Step four, check coverage. For each cluster, find the page that best serves it. If several pages overlap, pick the strongest and plan to merge the rest into it. If none exists, you have found a page to write. If one page covers several clusters poorly, plan to split it.
Step five, write or revise so the page answers the whole cluster, with the main question first, and variants used where they fit the sentence naturally. Keep the clustering sheet. It becomes your content map.
Merging overlapping pages safely
When two pages answer one need, keep the stronger one, fold the best material from the other into it and redirect the weaker address to the survivor. Done carefully, it consolidates your signals instead of losing them.
Choose the survivor by evidence: the page with more impressions, more links, a better address and the clearer answer. Copy anything valuable from the other page into it, such as a useful table, a unique example or a good question and answer. Do not simply paste both together, because the result reads as a bloated list rather than one complete answer.
Then set up a permanent redirect from the old address to the survivor, so visitors and links follow, and update your internal links to point straight at the new address. Our glossary covers the mechanics under 301 redirect. Submit the updated sitemap and watch Search Console over the next weeks. Expect some movement before it settles.
Do not merge pages that serve distinct intent just because the words overlap. Merge in response to real overlap, which you saw when checking the results and reading the pages.
If the pages were published recently and have no history, you can often just delete the weaker one with a redirect and lose nothing. The older and more linked a page is, the more care it deserves.
Why clusters matter for AI answers
AI assistants answer whole needs, including the follow-ups, so a page that covers a cluster completely gives them more to quote than a page that covers one phrasing. That is the second reason to build this way.
When an assistant answers "how long do shingles last," it may also search for what shortens their life and signs of wear, because a good answer includes them. A cluster page has those passages under their own headings, each one ready to be lifted. A single-keyword page has the first part only, and the assistant reads another site for the rest.
Clusters also make internal linking cleaner. Each cluster page links to its neighbors, so the structure shows what you cover. Our article on topic clusters for ecommerce shows how that works for a store.
Finally, clusters make measurement saner. When one page owns one need, you can see in Search Console whether it earns impressions across the phrasings, and whether an assistant names you when you ask the cluster's main question. For the cross-engine picture, see how to show up in AI search.
A worked example with a small business
Here is a cluster exercise for a fictional kitchen remodeling company, to show the steps with real-looking data. The business and its queries are invented for illustration.
The owner exports queries and finds these, each landing on a different page: "kitchen remodel cost," "how much to remodel a kitchen," "average kitchen renovation price," "kitchen remodel timeline," "how long does a kitchen remodel take," "kitchen remodel permits."
First pass by subject: all six concern kitchen remodeling. Second pass by intent: three are about price, two are about duration, one is about permits. That gives three clusters, not six pages.
Cluster one, what a kitchen remodel costs. Three phrasings, one need. One page covers the factors that drive cost, how to think about a budget and what a quote should include.
Cluster two, how long a remodel takes. Two phrasings, one page that walks through the stages in order and where delays happen.
Cluster three, permits. A distinct need, a distinct page, with local rules and who pulls them.
Before the exercise, the site had six thin pages. After it, three complete ones. Each can now open with a plain answer and cover its follow-ups. The owner links the three pages to one another and to the main remodeling service page.
The same exercise on your own data usually shows a handful of obvious merges and one or two missing pages. That list is your next quarter's work.
Signals that the change is working
Watch impressions per page, the number of distinct queries each page appears for, and how assistants answer the cluster's main question. Clicks lag behind these.
In Search Console, open a cluster page and look at the queries it now appears for. A healthy cluster page collects a growing list of related phrasings, which is the effect you were after. If it appears for only one phrasing, the page is probably still too narrow.
Watch for cannibalization: two of your pages swapping places for the same query. That means the merge is unfinished.
Then ask the assistants. Run the cluster's main question and its two or three follow-ups on a schedule and log whether you are named and what sources appear. Give it weeks, not days, and judge by the pattern.
If nothing moves, check the basics first: is the page indexed, does it open with the answer, and does it have links from other pages on your site? Clustering is a structure change, and it still needs the page itself to be good.
How RunOctopus handles this
RunOctopus is a done-for-you service. Its AI, Ollie, researches a business and then writes and publishes pages directly on the business's own website, with a team that plans, writes, reviews and publishes alongside it. It works with Shopify, WordPress, WooCommerce and Wix.
The planning is built around needs rather than single keywords: Ollie researches the questions your customers ask, groups the phrasings of one need together, and writes one thorough page for each, then links the pages to each other. Nothing to install, approve, upload or schedule, and pages go live on their own. The first week brings 8 long-form articles, 6 pages and a custom interactive tool built for your niche, and after that 175+ pieces of content a month.
RunOctopus also reports on where ChatGPT, Perplexity, Claude and Gemini name your business. The 90-day promise is more visibility on Google and in AI answers within 90 days, measured against the 90 days before you joined, or every dollar back.
To see how your site's pages are organized today, try the free site read at app.runoctopus.com/start. Paste your address and Ollie reads the site and shows the plan. No account, no call.