Keyword clustering is the step between a long list of keywords and a plan of pages. Ahrefs defines it as “the process of grouping keywords with the same or similar intent.” You collect hundreds of searches, then sort them into groups, and each group becomes a page or a section of one. This page is part of our guide to question keywords, where clustering is how a raw question list turns into something you can build.
Why cluster keywords at all
Because people ask for the same thing in many ways. Google said exactly this when it added query groups to Search Console in October 2025. Its post notes that people phrase the same need with “common misspellings, slightly different phrasing, and different languages.” Its example was the need behind “how to make guacamole dip?”, which also shows up as “guacamole dip recipe” and “guac dip recipe.” In Google’s words: “While these are distinct queries, they reflect a similar user intent.”
If you wrote a page for each of those, you would have three thin pages competing with each other for one need. A cluster gives you one strong page instead. Google’s helpful content checklist asks whether your content provides “a substantial, complete, or comprehensive description of the topic.” A page built from a whole cluster has a much better chance of doing that than a page built from one phrase.
It also saves you from yourself. Google’s generative AI guide says “AI systems can understand synonyms and general meanings of what someone is seeking.” You don’t need a page per wording for Google to connect your page to the need.
Three ways to cluster keywords
1. By meaning
Read the list and group what means the same thing. It is slow, but it is the most accurate method for small lists, and it is how you catch groups that share no words at all. “How tall is barack obama” and “barack obama height” look different and want the same answer.
2. By results-page overlap
This is the method most clustering tools use. Ahrefs describes it as “typically done by grouping keywords with the same or similar search results.” The idea is simple: if Google shows mostly the same pages for two searches, it treats them as the same need, so one page can rank for both. If the results are different, Google sees two needs, and you probably need two pages.
There is no official cut-off for how much overlap counts. Ahrefs’ own approach gives each pair a similarity score and keeps the rule loose: “high = cluster, low = don’t, middle = take your best guess.” Some tools compare the top 10 results for each keyword, some the top 100.
You can do a rough version by hand. Search both keywords in the same browser window, signed out, and compare the first page. Lots of the same URLs: one cluster. Few or none: keep them apart.
3. By People Also Ask parent
For question keywords there is a shortcut. When you open a question in People Also Ask, Google adds new questions under it. So every question you collect comes with a parent: the question that made Google show it. Questions under the same parent are already grouped, by Google, around one topic.
That doesn’t make the grouping final. Rewordings of one question often show up under different parents, and some children drift off topic. But it is a strong first draft. PAA Box, our Chrome extension, keeps that parent column on every row of its CSV export. The method for turning those groups into pages is in turning People Also Ask into a content plan.
How to cluster keywords, step by step
- Collect widely. Pull keywords and questions from several sources and seeds. Our guide to finding what people are searching for covers the free ones, and the long tail is where most of your candidates will come from (see long tail keywords).
- Clean the list. Remove exact duplicates and anything off topic. Mark searches that clearly have a different intent from the rest.
- Do a first pass by meaning. Group the obvious ones. Name each group with its clearest wording, ideally a question a person would ask.
- Check the borderline pairs on the results page. Whenever you aren’t sure two groups are the same need, compare their results. Merge or split based on what Google shows, not on how similar the words look.
- Split by intent. A group that mixes learning and buying is two groups. Our page on search intent covers how to tell them apart.
- Decide page or section. A big group with sub-questions of its own is a page. A small group that answers one part of a bigger topic is a section on that topic’s page.
- Arrange the pages. Put broad pages above specific ones and link them both ways. Google’s SEO Starter Guide recommends “using directories (or folders) to group similar topics,” which is the same idea applied to your URLs.
What a finished cluster looks like
A useful cluster has four parts:
- A main keyword. The clearest, most common wording. It goes in the title and the H1.
- Variants. Rewordings and misspellings of the same need. They don’t need headings. Use them naturally where they fit.
- Sub-questions. Related questions people ask next. These become H2s on the page.
- A home. The one URL that owns the cluster. Every other page that mentions the topic links there instead of answering it again.
That last part is what keeps your site from competing with itself. One need, one page.
Clusters and AI search
Clustering matters more, not less, as Google answers more searches with AI. Google’s generative AI guide describes query fan-out as “a set of concurrent, related queries generated by the model to request more information.” Its example: for “how to fix a lawn that’s full of weeds”, fan-out searches might include “best herbicides for lawns”, “remove weeds without chemicals” and “how to prevent weeds in lawn”.
Those are the sub-questions of a cluster. A page that answers the main question and its likely follow-ups is the kind of page those related searches find. The same guide warns that creating content for fan-out queries mainly to manipulate rankings breaks Google’s scaled content abuse policy, so answer the follow-ups because readers need them, on one page, rather than spinning out a page for each.
Keyword clustering tools
- A spreadsheet. For a few hundred keywords, a column for the cluster name and some sorting is enough.
- Results-overlap tools. Several SEO tools cluster by comparing search results automatically. Ahrefs names Keyword Insights as one that groups keywords “by comparing the top 10 or 100 search results.” Check what each tool compares and how it scores before you trust its groups.
- Search Console query groups. If your site gets a lot of searches, Search Console Insights may group your queries for you. Google says the feature is “available only to properties that have a large volume of queries,” and that the groups “are computed using AI; they may evolve and change over time.” It works on searches you already appear for, which makes it a good check on clusters you have already built.
For free keyword research tools in general, see free keyword research tools.
Common mistakes
- Clustering by shared words. “Coffee scoop” and “coffee scoop size” share words but may want different things: one a product, one a measurement. Check the results page.
- Clusters that are too big. If a cluster needs ten H2s that each deserve their own answer, it is a hub with several pages under it, not one page.
- One page per variant. The opposite mistake. Variants of one need share a page.
- Mixing intents. Learning and buying rarely belong on the same page.
- Clustering once. Results change. Re-check your most important clusters when rankings or results shift.
Where to go next
Go back to question keywords for the full process. To see which questions in a topic come up most, read the most asked questions on Google. And if your source is People Also Ask, the step-by-step method is on people also ask keyword research.