Keyword Clustering by SERP Overlap: How It Works, Thresholds and Tools (2026)

1What is keyword clustering by SERP overlap?

Keyword lists grow quickly, and the question that matters is how many pages they need. Clustering answers it by grouping keywords that belong on the same page. Overlap clustering does this with one observable fact: the URLs Google ranks. If the top ten results for two keywords share enough URLs, Google is treating them as one need, and a single page can target both. If they share almost nothing, they are different needs, and one page will rarely serve both.

The steps are the same everywhere. Fetch the top ten for every keyword. Count the URLs each pair has in common. Group the pairs that meet your threshold. Name each group by its head keyword, usually the one with the most demand. Then assign one target page per group. Nightwatch describes the measure the same way: "counting how many URLs the two SERPs share," with "a common threshold" of 3 to 4 overlapping URLs in the top ten (Nightwatch).

2What does SERP overlap look like on real results?

Both examples below come from live Google results we fetched on 28 October 2026 (United States, English, through our SERP scraper). Our scraper returned 9 organic results for the first keyword and 8 for the second, so the overlap is counted on those, not on a full ten each. The table lists every URL from either top list.

Overlap count: 3. Three URLs (seo.ai, Zenbrief and thruuu) appear for both keywords. With a threshold of 3, the two keywords form one cluster. With a threshold of 4, they split. That is why the threshold is the decision that matters, and it is the honest reason to look at the pairs sitting right at your cut off rather than trusting the tool.

The second check is the one wording cannot make. "Keyword clustering tool" and "keyword grouping tool" mean the same thing to a person. Their results shared zero URLs in the lists we fetched: the first returned roundups and tool pages such as lowfruits.io and Semrush, the second returned a different set led by tenscores.com and an SE Ranking roundup. A wording based grouper would put them together. An overlap grouper keeps them apart, and tells you that ranking for one will not rank you for the other.

3Why does SERP overlap beat grouping by wording?

Grouping by wording or by meaning is a guess about what searchers want. Overlap uses Google's answer. When similar phrases return a guide for one and product pages for the other, only the results reveal it. That is also why the AI Overview on the "keyword clustering" results page describes the contrast as semantic clustering grouping "words by natural language meaning" while SERP clustering "relies on real search engine behavior."

Semantic and text based clustering still have a place. They need no live results, so they are cheap and fast on very large lists, and they work for keywords no one ranks for yet. The tools table below shows the split: Zenbrief uses embeddings and Pemavor uses text vectors, while the rest compare results. A practical hybrid is to use a semantic pass to cut a huge list down, then overlap to settle which keywords share a page. Nightwatch recommends combining intent, semantic and SERP methods, and still wants a human review of overlap before publishing.

4How do you cluster keywords by SERP overlap by hand?

You can do this in a spreadsheet for a short list. Search each keyword in the same market and language, copy its top ten URLs into one column per keyword, then compare columns. For each pair, the overlap is the number of URLs that appear in both, which a COUNTIF across the two columns gives you. Group pairs at or above your threshold, and check that every member of a group overlaps with the head keyword.

The limit is arithmetic. Comparing every keyword with every other one needs n(n-1)/2 comparisons, so 100 keywords is 4,950 pairs and 300 is 44,850. Past a few dozen keywords the spreadsheet stops being sensible. That is the case for tools, and for the soft method some tools offer, which compares each keyword only with the largest one.

5What overlap threshold should you choose?

The threshold is the minimum number of shared top ten URLs that puts two keywords in one cluster. The trade-off is documented by SE Ranking: a higher setting means fewer keywords land in each group and more stay ungrouped, while a lower one makes broader clusters that you will probably need to regroup by hand (SE Ranking). Put plainly:

• Too low (1 to 2): unrelated keywords chain together through a single shared page, and you end up with clusters that mix intents and force one page to do two jobs.
• Too high (5 and above): near synonyms split apart, you plan more pages than you need, and those pages end up competing with each other.
• Starting point, 3: the fixed rule at KeyClusters, the setting Linkeddit Topics uses, and the low end of the 3 to 4 range Nightwatch calls common. Treat it as a default to test, not a proven optimum: the tools we checked publish settings, not accuracy studies for them.

Our recommendation is to cluster at 3, then read the pairs that scored exactly 2 and exactly 3, since those are where the decision is close. Look at what kind of page ranks, not only how many. If the shared URLs are the same article type and the same angle, merge. If they are a forum thread, a product page and a guide, the pair shares a topic but not an intent.

6Which keyword clustering tools use SERP overlap?

Every row below was read on the tool's own page (or, for Semrush and Keyword Insights, its own blog) on 28 October 2026. We omit what a page does not state. Prices change, so treat them as a dated snapshot.

Sources: SE Ranking, Semrush, ContentGecko, KeyClusters, Keyword Insights, Zenbrief, Pemavor.

Two points the roundups that rank for this query rarely make. First, none of the pages we read documents an accuracy figure for its threshold, so a claim that one tool is "most accurate" rests on the reviewer's judgement. Second, the tools differ in whether the threshold is yours to set: KeyClusters fixes it, SE Ranking and ContentGecko expose it, and Semrush documents none. If you want to test settings, choose one that lets you re-cluster the same list at different thresholds, as ContentGecko's paid tier describes without spending credits again.

7Linkeddit Search Topics: one option (ours)

Disclosure: this is our product. The Topics view in Linkeddit Search clusters by SERP overlap on keywords you already track. Each checked keyword has its top ten results stored, and keywords that share at least three of those results form one topic, named by a head keyword. A run starts from your active tracked keywords and your top Search Console queries and covers 300 keywords by default.

Each topic carries one action, which is where it differs from a plain grouper:

• Striking distance: your best rank is 11 to 20.
• Gap: no page of yours in the top 20 while a competitor is in the top 10.
• Cannibalised: two or more of your pages rank.
• Protected: you rank 1 to 3.
• Hold: nothing is flagged.

One limit to know: each check reads only the first ten results, so a rank of 11 to 20 is not recorded from checks alone, and a topic whose best page sits there reads as Gap. Cost is shown first: a run opens an estimate panel, nothing is spent until you press Start run, and keywords whose results were checked in the last 7 days are free. It is part of the $60 per month Compete plan. Setup is in the Search keyword research guide, and the tracking behind it is on the rank tracker page. It suits you if you already track keywords with us. If you want to cluster a list of keywords you have never checked, or choose your own threshold, a tool in the table above fits better.

8How do clusters become one page per topic, and expose cannibalisation?

Cluster output is a list of groups. The work is deciding what each one means for your site. Compare each cluster against the pages you already have, then apply the rule that fits. This is our guidance, not a vendor standard:

Cannibalisation is the row that matters most, and clustering finds it from the other direction. Instead of starting from a page and asking what it ranks for, you start from the cluster and ask which of your URLs show up in it. Two of your pages in one cluster means Google is choosing between them for the same need. Semrush defines it as multiple pages that "target the same keyword(s) and harm each other's search visibility," and adds the useful caveat that several pages ranking for a query is not always cannibalisation, especially for branded searches where each page serves the intent (Semrush).

To confirm a suspected case, use Search Console: filter by the query, open the Pages tab, and see whether more than one URL earns impressions and clicks. To fix it, Semrush's guide has you designate one preferred page and signal it with a redirect, a canonical tag, or by differentiating each page, keeping noindex as a last resort because it consolidates nothing. Prevention is cheaper: map every keyword to one URL before you publish. To check whether a merge helped, see did my SEO change work.

9What are the cost, freshness and top ten limits?

• Top ten only. Overlap sees ranks 1 to 10. A page sitting at 11 to 20 is invisible to it, and tools that read only the first ten cannot tell you a page is close. Check ranks beyond ten in Search Console.
• Freshness. Results change, as the two fetches we took minutes apart showed. Nightwatch warns that a cluster valid 18 months ago may have fractured. Re-check before you merge or write.
• Cost scales with keywords, not pairs. A tool needs one results fetch per keyword, which is why pricing is per keyword: KeyClusters charges one credit each, and Linkeddit counts checks against a monthly allowance, with recently checked keywords free.
• Market and device. Results differ by country, language and personalisation. Cluster in the market you sell in, and do not reuse a US cluster for another country.
• Small lists chain badly. In a niche with few strong pages, the same URLs can appear everywhere and a low threshold may merge things that are not alike. Raise it or review by hand.

Frequently asked questions