Keyword Clustering: How to Group Keywords for SEO (and AI Search)

You run keyword research and end up with 800 terms that all say roughly the same thing. Build a separate page for each and you get thin content, pages competing with each other, and hours lost. Keyword clustering fixes that. You group terms that share the same search intent, then target each group with one strong page. This guide walks you through what clustering is, the methods and tools behind it, and a step-by-step process you can run today.

What is keyword clustering?

Keyword clustering is the practice of grouping search terms that share the same search intent so a single page can target the whole group. Instead of writing one page per keyword, you cover a cluster of related terms with one thorough article. This lets that page rank for many queries at once.

Every cluster has a primary keyword, the main term you want to rank for, and secondary keywords, the synonyms and long-tail variants around it. Take “king size mattress”, “king mattress” and “king bed size”. A searcher wants the same thing each time, and Google shows near-identical results, so all three belong on one page.

The payoff is combined search volume. On their own, many of those variants have low volume. Grouped onto one page, their volumes add up, and you chase that traffic with a single piece of content instead of five.

	SERP-based clustering	Semantic (NLP) clustering
Groups by	Shared ranking URLs in live results	Meaning and word similarity
Reads intent from	Google itself	A language model’s interpretation
Best for	Accuracy on what ranks together	Speed on very large keyword lists

Why keyword clustering matters for SEO

Clustering is not busywork. It changes how well a whole site performs, for four reasons.

Stops keyword cannibalisation

When you publish separate pages for “email marketing tips”, “email marketing advice” and “email marketing best practices”, they compete for the same rankings. Google has to pick one, splits signals between them, and often ranks none of them well. That is keyword cannibalisation: your own pages eating each other’s chances. Clustering those terms onto one page removes the internal competition.

Builds topical authority and E-E-A-T

A single page that answers every related question shows a search engine you cover the subject in depth. That depth signals expertise, one of the pillars Google groups under E-E-A-T (experience, expertise, authoritativeness and trustworthiness). Cover a topic thoroughly across a cluster, and you build the kind of authority that lifts rankings for the whole theme, not just one term.

Captures combined search volume with fewer pages

Long-tail variants often draw only a handful of searches each. Ten of them on one page, though, can add up to real monthly traffic. You capture that combined demand with one article, which is far less work than maintaining ten thin posts that each chase a single term.

Shapes site structure and internal linking

Clusters give you a blueprint for your site. Each cluster maps to a page, and related clusters map to sections, so your categories, subcategories and individual posts fall into a clear hierarchy. That structure makes internal linking obvious: you link spokes to their pillar and back, which helps both readers and crawlers move through your content.

Keyword cluster vs topic cluster vs keyword grouping

People use these three terms as if they mean the same thing. They do not, and mixing them up leads to messy content plans.

A keyword cluster is a set of same-intent search terms you target with one page. A topic cluster is bigger: one pillar page on a broad subject, linked to many supporting pages (spokes) that each cover a subtopic. Keyword grouping usually refers to the older method of sorting keywords by shared words rather than by what actually ranks. The table below sums up the differences.

Term	What it groups	Scope	Output
Keyword cluster	Same-intent search terms	One page	A single optimised page
Topic cluster	Pillar plus subtopics	Many pages	A linked pillar-and-spoke structure
Keyword grouping	Terms with shared words or stems	Varies	A sorted keyword list

The methods behind clustering: SERP-based vs semantic vs lexical

Tools cluster keywords in three broad ways. Knowing which method a tool uses tells you how much you can trust its groups, so it is worth understanding all three before you pick one.

SERP-based (SERP overlap) clustering

SERP-based clustering compares the live search results for each keyword. The tool pulls the top-ranking URLs for two terms and checks how many they share. If enough of the same pages rank for both, the terms go in one cluster. The logic is simple: if Google already ranks the same pages for two queries, Google treats them as the same intent, so one page can serve both. This method reads intent straight from the search engine, which makes it the most accurate for deciding what can share a page.

Semantic (NLP) clustering

Semantic clustering groups keywords by meaning rather than by results. It uses natural language processing, the same branch of machine learning behind Google’s own BERT language model, to judge how closely two phrases relate in meaning. It is fast and cheap, and it handles huge lists well. The catch is that two phrases can mean almost the same thing to a language model yet return very different search results, so semantic groups sometimes miss the real intent that a live SERP would reveal.

Lexical / n-gram / lemmatisation clustering

The oldest method groups keywords by the words themselves. Lemmatisation reduces each word to its base form (so “running”, “runs” and “ran” all become “run”), then groups terms that share those base words. An n-gram approach does something similar using shared word sequences. This is quick and needs no live search data, but it is the bluntest of the three: it groups by spelling, not by what searchers actually want, so it misses same-intent terms that use different words.

The concept and the first SERP-based algorithm were introduced in 2015 by Alexey Chekushin, a Russian SEO specialist, who also defined the soft and hard clustering types covered next. Before that, lemma-based grouping was the norm, which is why SERP-based clustering felt like such a step forward.

Understanding the clustering level (soft, hard and complete)

Here is the lever most guides skip, and the one that decides whether your clusters are useful or a mess. Every SERP-based tool lets you set a clustering level, and getting it right matters more than which tool you use.

What the clustering level (threshold) controls

The clustering level is the minimum number of shared URLs two keywords need before a tool groups them. Most tools compare the top 10 results, so the level is a number from 1 to 10 (some express it as a percentage of overlap).

The level works as a dial. Set it low, say 2 or 3 shared URLs, and you get fewer, larger clusters that cast a wide net but can bundle loosely related terms together. Set it high, say 7 or more, and you get more, tighter clusters where Google’s results are nearly identical. Most practitioners start around 3 to 4 shared URLs and adjust from there. If a cluster looks too broad to cover on one page, raise the level and re-run it.

Soft clustering

Soft clustering is the loose setting. Two keywords join a group if they share at least one URL in their top results. Some tools also use “soft” to mean a keyword can sit in more than one cluster at once, which helps for queries that straddle two intents. Either way, soft clustering casts the widest net, so expect to review the groups by hand.

Hard clustering

Hard clustering is the strict setting. Keywords group only when they share most or all of their top-ranking URLs, so the results are tight and reliable. In a hard setup, each keyword lands in exactly one cluster, with no overlap between groups. Use it when you want confidence that everything in a cluster genuinely belongs on the same page.

Complete clustering

Complete clustering is the tightest form of all. Here every keyword in a group shares matching URLs with every other keyword in that group, not just with the group’s main term. That produces very small, very precise clusters. It is overkill for most projects, but useful when you need near-certainty that a set of terms shares one intent, such as mapping money pages where a wrong grouping is costly.

How to do keyword clustering: a step-by-step process

You now know the theory. Here is the practical run, seven steps from a raw keyword list to a page that ranks. Work through them in order.

Step 1: Build your keyword list

Start with a few seed keywords, the broad terms that describe your business, then expand them into a full list. A keyword tool such as SE Ranking or the Keyword Magic Tool will suggest hundreds of related terms. Pull the search volume, keyword difficulty and intent for each, and (this matters for later) the top-ranking URLs. Add a rival’s ranking keywords too, to fill gaps you would not think of on your own.

Step 2: Classify search intent

Before you group anything, tag each keyword with its intent, because keywords with different intent almost never belong on the same page. There are four main types of search intent: informational (learning something), navigational (finding a specific site), commercial (researching before buying) and transactional (ready to act). Modifiers give you a fast first read: “how” and “what” signal informational, “best” and “vs” signal commercial, “buy” and “price” signal transactional. When a modifier is unclear, the live SERP settles it.

Step 3: Group by SERP overlap and intent

Within each intent bucket, group the terms whose search results overlap. You can do this by hand for a short list: search your top keyword, note the top 10 URLs, then check each other keyword against it and group the ones that share enough results. For anything above 100 keywords, a tool does this far faster. Either way, ask three questions of a borderline group: do the same pages rank for these terms, could one page realistically cover them all, and would a searcher want them answered together?

Step 4: Assign each cluster to a page (new or existing)

Every cluster maps to one page. Before you create a new one, check whether an existing page already targets that cluster. If it does, optimise the existing page rather than publishing a competitor to it, which is how cannibalisation creeps back in. If nothing fits, plan a new page and slot it into your site structure as a pillar or a spoke. This is where local SEO clusters, for instance, often map to their own dedicated pages.

Step 5: Prioritise clusters

You cannot build every page at once, so rank the clusters. Weigh combined search volume, keyword difficulty, and business impact together, not volume alone. A smaller cluster with strong commercial intent often earns more than a high-volume informational one. Factor in effort, too: refreshing an existing page is quicker than writing a new one from scratch, so some quick wins are worth doing first.

Step 6: Write or optimise the page

Write one thorough page for the cluster. Put the primary keyword in the URL, title tag, H1, meta description and first paragraph, then work the secondary keywords naturally into subheadings and body copy. Do not force them: if a term will not fit cleanly, leave it out. Cover the subtopics your cluster implies so the page genuinely answers the whole group, not just the main term.

Step 7: Track cluster rankings

Publishing is not the end. Track how the page ranks for the whole cluster, not just the primary keyword, using Google Search Console or a rank tracker. Watch for terms that underperform, which can mean the cluster was too broad and needs splitting, or that a different page is competing. Re-check every few months, since SERPs shift and a cluster that held together last year may not today.

Manual vs automated clustering: which to use

The choice comes down to list size and how much control you want. Manual clustering, where you compare SERPs yourself, gives you full control and a real feel for each group, but it is slow. Doing a thousand keywords by hand can take days. Automated tools cluster hundreds or thousands of terms in minutes, at the cost of some judgement you then have to add back.

A simple rule works well. For a list under about 100 keywords, or for high-value money pages where a wrong grouping is costly, cluster by hand. For anything larger, run a tool, then review the output rather than trusting it blindly. Many teams combine both: let a tool draft the clusters, then regroup a few by hand to fit their site. You get the speed of automation and the accuracy of a human check.

Keyword clustering tools compared

Most SEO platforms now include a clustering feature. Here are five worth knowing, from full suites to lightweight options, so you can match one to your list size and budget.

SE Ranking keyword grouper

SE Ranking’s grouper uses SERP-based clustering and returns groups with volume, difficulty and cost-per-click for each term. It shows clusters as a mind map, with your primary keyword in the centre and related terms branching off, which makes the topic structure easy to read at a glance. It suits agencies that already track rankings in SE Ranking and want clustering in the same place.

Keyword Insights

Keyword Insights is a dedicated clustering and intent platform. It examines the top search results for each keyword and groups terms that share a set percentage of URLs, a threshold you can adjust to make clusters tighter or looser. It also classifies intent and pulls your current rankings, so you can filter for clusters where you have no content yet. Those are your content gaps, surfaced automatically.

Keyword Cupid

If you searched for how to cluster keywords with Keyword Cupid, here is the short version. You upload a keyword list as a CSV, and the tool scrapes live Google results for every term, then trains a machine learning model to group keywords by SERP overlap. When two or more keywords share ranking URLs, they go in one cluster; when they share none, they split apart. You get the results as an interactive mind map you can drill into. It leans on live SERP data rather than word matching, so its groups reflect what Google actually ranks.

Semrush Keyword Strategy Builder

If you already pay for Semrush, its Keyword Strategy Builder clusters a keyword list automatically. Open a list, click “Cluster this list”, and it groups terms into a pillar page with supporting subpages, each with a primary and secondary keywords. You can then send a cluster straight to Semrush’s writing tool. It is convenient for existing subscribers, though it needs a paid plan.

Free and lightweight options

You do not always need a paid suite. Free or low-cost tools such as an n-gram grouping tool cluster small lists by shared words, and simple SERP-overlap checkers handle a few hundred keywords. There are also free AI generators and downloadable templates for one-off jobs. These may be a good choice if you cluster occasionally or work with short lists, though they lack the intent data and scale of the paid platforms.

Keyword clustering for AI search (GEO / AEO)

Clustering is no longer only about the traditional blue links. AI-powered search, including Google’s AI Overviews, ChatGPT, Perplexity and Gemini, changes what gets seen, and clustering feeds it directly. Optimising for these systems is sometimes called generative engine optimisation (GEO) or answer engine optimisation (AEO).

Here is why it matters. These systems use a technique called query fan-out: they break one prompt into several related sub-queries, then pull answers from pages that address them. Pages that answer multiple related sub-questions in one place give the AI more reason to cite them. That is exactly what a well-clustered page does. By covering a full group of related terms, it lines up with the fan-out queries an AI generates, which makes it more likely to be quoted in an answer.

“Keyword clustering chatgpt” is a rising search for a reason. You can paste a keyword list into a chatbot and ask it to group the terms by intent, and for a quick first pass it works. But a language model guesses at intent from meaning alone, the same weakness as semantic clustering, so it misses cases a live SERP would catch. Treat an AI’s clusters as a draft to validate against real search results, not a finished map. If you want to build genuine AI search visibility, our AI SEO services go deeper into this.

Common keyword clustering mistakes to avoid

Clustering is simple in principle and easy to get wrong in practice. These five mistakes trip up most people, so watch for them.

Clustering by word similarity instead of intent

Two keywords can share almost every word yet mean different things. “Apple cider vinegar for dog shampoo” (a how-to) and “apple cider vinegar dog shampoo” (a product) read alike but want different pages. Group by intent and by what actually ranks, not by matching words, or you will build one page that serves neither searcher well.

Setting the clustering level wrong for the list

Leave the threshold on its default and you may get clusters that are too broad to cover on one page, or so tight that near-identical terms split into separate groups. Check your first output, then adjust the level and re-run. The right setting depends on your list, so treat the first pass as a draft.

Forcing unrelated keywords into one page

A big cluster looks efficient, but cramming loosely related terms onto one page dilutes it. If a group covers two genuinely different questions, split it into two pages. One page should answer one intent well, not two intents poorly.

Ignoring existing pages and creating the cannibalisation you meant to fix

Clustering is supposed to end pages competing with each other. Publish a new page for a cluster you already cover, though, and you recreate the problem. Always check your existing content first, and optimise what you have before adding something new.

Never re-checking clusters after the SERP shifts

Search results change. A cluster that held together last year may not today, as Google reshuffles what ranks. Set a reminder to re-run your important clusters every few months, and split or merge them as the SERPs move.

A worked example: clustering for a Singapore business

Theory sticks better with a real example, so here is a short one. Say you run an aesthetic clinic in Singapore and you have pulled this small keyword list: sculptra singapore, sculptra treatment, sculptra price singapore, collagen stimulator injection, best moisturiser for dry skin, and how to fix sagging skin.

First, tag each by intent. The first four are commercial or transactional (someone researching or ready to book a treatment). “Best moisturiser for dry skin” is commercial but sits in a different product world. “How to fix sagging skin” is informational.

Next, check what ranks. “Sculptra singapore”, “sculptra treatment”, “sculptra price singapore” and “collagen stimulator injection” all pull up the same treatment pages, so they form one cluster around a single Sculptra service page. “Best moisturiser for dry skin” ranks alongside product roundups, not clinics, so it splits off. “How to fix sagging skin” returns guides, so it becomes a blog post that links to your Sculptra page.

You end up with three pages, each mapped to one intent, instead of six thin pages fighting each other. That is clustering doing its job. If you sell online rather than treat in a clinic, the same logic maps product and category keywords onto Shopify SEO pages.

Frequently Asked Questions

Final thoughts

Keyword clustering turns a sprawling keyword list into a clear plan: fewer, stronger pages that each own an intent, rank for more terms, and stand a better chance of being cited in AI answers. Start small. Take your next keyword list, tag each term by intent, and group the ones that share search results. You will spot the overlaps and cannibalisation almost straight away.

If you would rather have this done properly across your whole site, our SEO agency in Singapore build clustered content strategies from the ground up. Found this useful? Share it with someone wrestling with a messy keyword list, and tell us in the comments how you cluster yours.