Keyword Clustering: A Guide to the Free SEOquick Grouping Tool

Anyone can collect keywords today: a Google Search Console export, Serpstat or Ahrefs reports, Google Keyword Planner — there is no shortage of sources.

But then comes the problem almost every SEO course stays silent about: how do you distribute those keywords across pages?

If you group queries incorrectly, you will end up promoting irrelevant pages. Google will show a different page than the one you optimized for a query, rankings will keep “jumping,” and six months of work will go down the drain. I have seen this dozens of times — even with experienced specialists.

In this guide I will break down every working clustering method, show where ChatGPT genuinely helps (and where it fails), and walk you through every setting of our free tool.

Keyword clustering is the grouping of search queries by meaning and intent so that each group (cluster) corresponds to one page of a website. Keywords can be grouped in three ways: by matching URLs in Google’s search results (SERP-based clustering), by word composition (the lexical method used in the SEOquick clustering tool), and manually. Without clustering, a keyword list is just a pile of queries you cannot build a site structure on.

Where to Get Keywords in 2026

Before grouping, you need to collect. Here are the sources we use at our agency:

1. Google Search Console — the most underrated source. These are real queries your site already appears for. Export the Performance report to CSV or Google Sheets. Since December 2025, GSC supports natural-language filters: you can type “show queries with high impressions and low CTR over 28 days” — and the console builds the report itself. A perfect base for expanding the semantics of existing pages.
2. Google Keyword Planner — official Google data on search volume plus suggestions for new keywords. Without active ad campaigns it only shows ranges, but that is enough for a rough cluster estimate.
3. Serpstat — your site’s and competitors’ rankings, search suggestions, related phrases. For Ukraine and Eastern European markets, one of the most complete indexes.
4. Ahrefs — competitors’ keywords, related terms, questions. The strongest database for Western markets.
5. Our free keyword generator — quickly combines word forms and builds query variations.
6. Third-party services like keywordtool.io — they scrape Google, YouTube, and Amazon suggestions.

We merge all exports into a single file — and get a “heap” of thousands of queries that needs to be turned into a site structure.

Exporting from Google Search Console takes a couple of clicks: open the Performance report, set the date range, and click “Export → Google Sheets” (or CSV). On the “Queries” tab you get a simple table — the first two columns are what we care about:

The “Query” column becomes the basis of your semantics; the rest helps you gauge potential.

Clustering Methods: by SERP, Manually, by Words

Method One: SERP-Based Clustering

The most “scientific” approach is to trust Google itself. The logic is simple:

1. Take a list of keywords.
2. Pull the top 10 search results for each of them.
3. Compare how many URLs overlap between the SERPs of two queries.
4. If there is enough overlap, the keywords go into the same cluster.

The standard threshold is 3-4 shared URLs in the top 10. Hence the familiar modes in clustering services:

Hard produces small but very precise groups; Soft produces large ones, with the risk of “hooking” something irrelevant.

What is the catch with the SERP method? Search results are unstable. Semantics clustered two months ago may group differently today — the sites in the top 10 have changed, especially after another Core Update. Imagine an SEO specialist who reshuffles keywords back and forth every three months with no justification whatsoever.

Then there is the economics: pulling SERPs costs money, and with paid services clustering 1,000 keywords averages a few dollars. When I was building semantics for a new site, I had to process 80,000 queries from different sources. Paid SERP clustering at that volume would have cost hundreds of dollars — for nothing.

And the main issue: the algorithms are closed. The service will not explain why obvious synonyms ended up in different clusters, and there are usually no settings to fix it.

Method Two: Manually

Okay, you decided not to pay. You have Excel, patience, and time.

Take a mind map, write out all your services and products, and design the structure:

• Homepage — the core business activity. An SEO agency, an online clothing store, a dental clinic in Kyiv — this is where the highest-volume keywords live.
• Top-level categories — fundamentally different groups of products/services. An electronics store does not mix cameras with smartphones. The list is built through brainstorming and competitor analysis.
• Subcategories and landing pages — this is where 99% of the semantics you promote ends up. Each page should hold traffic-driving keywords united by a single logic.
• Product/service pages — keywords are mapped here less often; the exception is optimizing product names for high-volume keys.
• Blog — all informational and question-type queries. Do not throw them away: collect them into a separate pool and sort them by topic.

Two tips from practice:

Tip 1. Check phrase volumes in broad and phrase match — via Keyword Planner, Serpstat, or Ahrefs. Do not work with “dead” keywords.

Tip 2. Look for synonyms: Serpstat and Ahrefs have related-phrase reports, and a simple trick also works — type your main keyword into Google and study competitors’ snippets. You will immediately spot 2-3 alternative wordings.

The downside of the manual method is obvious: an average site takes 10-50 hours of monotonous eyeball regrouping. That is exactly why we built our own tool.

Method Three: by Word Composition (Our Method)

Lexical clustering groups queries by shared words and lemmas: “buy autoclave,” “autoclave buy Kyiv,” and “autoclave price buy” will land in one group without pulling any SERPs.

The advantages: instant, free, reproducible (the result does not depend on what is in the top 10 today), and fully controllable — you define the synonyms, negative keywords, and inseparable phrases yourself. The downside: the method does not see intent through Google’s eyes, so borderline groups should be spot-checked against the SERP. At large volumes (tens of thousands of keywords) it is the most rational balance of speed and quality.

SEOquick experience. Proper clustering is not theory. For an autoclave store we mapped the semantics across categories and product pages so that the site grew from 5,000 to 20,000 visits per month and took the #1 spot for “buy autoclave” (in the local market), beating the marketplaces — case study in our portfolio. And on a binary options project, clustering long-tail semantics became the foundation for 50,000+ programmatic pages and an 11x traffic increase: from 130 to 1,480 users per day — case details.

Clustering vs LLMs: Can You Group Keywords with ChatGPT?

The honest question of 2026: why bother with clustering tools at all when there are ChatGPT, Gemini, and Claude?

The short answer: an LLM groups well by meaning, but it does not see Google’s search results. And the final call — “one page or two” — is made by the SERP.

What an LLM does well:

• Instantly sorts a few hundred keywords by intent: commercial / informational / navigational.
• Understands synonyms and typos better than any lexical algorithm: it will merge “laptop for work” and “notebook for the office” without any synonym dictionary.
• Generates cluster names, H1s, and draft Titles for ready-made groups — routine work that used to be done by hand.
• Helps design the structure of pillar pages and supporting articles.

Where an LLM fails:

• It groups phrases that are “similar in meaning,” while Google groups by “similar SERPs.” These are not the same thing: semantically close queries can have completely different top-10s, and then one shared page will not rank for either. In a Keyword Insights test (2025), ChatGPT scored only 47/100 on clustering accuracy — a verdict for tasks where the cost of a mistake is six months of promotion.
• On large lists (5,000+ keywords) the model loses context: it duplicates clusters, “forgets” some phrases, and produces a different result on a repeat run.
• An LLM confidently hallucinates search volumes. Never ask ChatGPT “how many people search for this query” — the numbers will be made up.

A working prompt for intent grouping (for lists up to ~300 keywords):

You are an SEO specialist. Here is a list of keywords (one per line).
1. Determine the intent of each: commercial, informational, navigational.
2. Group the keywords by meaning into clusters; a cluster contains queries
   that ONE page can answer.
3. For each cluster, suggest a name, a page type (category / product page /
   article), and a draft H1.
Output as a table: cluster | intent | page type | keywords | H1.
Do not invent search volumes. If a keyword does not fit any cluster,
move it to an "Other" group.
[KEYWORD LIST]

Bottom line: in 2026 the working setup is a hybrid. A clustering tool (word-based or SERP-based) does the base grouping, and an LLM works on top of the ready groups — labeling intents, generating headings, and drafting copywriter briefs. We cover how to build this combo in detail in our articles on 50 mega-prompts for ChatGPT and Gemini in SEO and ChatGPT for SEO. And if you also want your pages to be cited in Google’s and ChatGPT’s AI answers, read our guide to GEO: optimizing your website for GPT.

▶ SEO-аналитика с ИИ делает работу 5 часов за 5 минут

Step-by-Step: The SEOquick Keyword Grouping Tool

We built a free tool that groups tens of thousands of keywords in minutes — with no SERP scraping and no per-thousand-rows fees.

We won’t show interface screenshots — the tool is being updated right now, and it’s faster to open it once and follow the steps live. Just open the clustering tool in a new tab, paste your keywords into the list column, and configure the fields for your project as you read. Let’s go through each one.

Treat as One Word

Semantics often contain inseparable phrases: iPhone 17 and iPhone 16 Pro are formally different phrases, and if you do not “glue” them together, every generation will land in one group. The same goes for car models written either as one word or with a space (A4 and A 4). List such combinations separated by commas — and the algorithm will treat each as a single word.

Negative Keywords

Exports from Serpstat, Ahrefs, or GSC always contain junk: “free,” “DIY,” other people’s brands. Paste your negative keywords separated by commas — and they will be removed from the list before clustering.

Exact word forms are supported: add an exclamation mark (!free) — and the system will remove only that exact form, keeping derivatives like “freely.” This mechanism will also be appreciated by anyone cleaning up keyword lists for PPC campaigns.

Ignored Words List

Prepositions, pronouns, and function words occur frequently, but they do not deserve a separate page: the groups “website development” and “development of websites” are essentially one. Add such words to the ignored list — and the algorithm will disregard them during grouping. English and Russian prepositions are built in by default.

Required Words List

The reverse situation: you scraped a competitor’s keywords, their product range is wider than yours, and the file contains thousands of phrases outside your niche. Fill in the “Required words” field (words or phrases separated by commas) — and only queries containing them will enter the clustering.

Synonyms

The most common failure of clustering tools is the inability to merge synonyms: “promotion,” “optimization,” and “ranking” of a website drift into different groups, which then have to be glued back together by hand. In our tool, synonyms are defined explicitly: list the variants on one line separated by commas; press “+” for a new synonym group. You can enter everything at once or keep adding as you work on the project.

All of these fields — negative, ignored, and required words plus the synonym rows with the “+” button — sit in a single settings panel. Rather than study them in a picture, open the clustering tool and walk through the fields yourself: drop in a couple of your own negative keywords and synonyms and watch the result change instantly.

Geo-Dependency Handling

If your semantics follow the pattern “every keyword + city,” rough clustering will throw two or three different cities into one group. But someone searching for “pizzeria Kyiv” and someone searching for “pizzeria Lviv” are different customers with different delivery conditions. The tool has a built-in database of cities in Ukraine and Europe, and a dedicated algorithm prevents different cities from landing in the same cluster. The feature is on by default.

Extended Clustering for Short Queries

Strict grouping often leaves short high-volume phrases “overboard.” This option checks unsorted short queries for the best match with a high-volume group and adds them to it. A feature for experienced users — off by default, enabled with a checkbox.

Clustering Strength

The principle is the same as Soft/Hard in SERP services, except the threshold is the number of matching words in a group:

1. HV (Soft) — a two-word match. Broad groups for the initial structure and commercial pages.
2. MV (Moderate) — a three-word match. Narrow groups for copywriter briefs and blog articles.
3. LV (Hard) — a four-word match. Dense clusters for large keyword sets and informational keywords.
4. Micro-volume — a five-word match. For content sites with massive long-tail semantics.

Grouping and Filters

Scrolling through thousands of result rows is inconvenient, so we added Excel-style filters: click the word “GROUP,” uncheck the groups you do not need, check the ones you do — and save the selection as separate files. Handy for preparing briefs for several pages at once. Run your own list through the clustering tool and try this filter on live results — it’s clearer than any screenshot.

Importing and Saving a Project

The project with all its settings is saved to an Excel file via the “SAVE” button. In that same file you can add more keywords, search volumes, competition, and cost-per-click data (for example, from Serpstat, Ahrefs, or Keyword Planner), as well as fill in every field: synonyms, negative, required, and ignored words. Then upload the file back — and continue right where you left off.

Imports of Serpstat, Ahrefs, and Keyword Planner exports are supported: the system recognizes the “keyword,” “volume,” “cost per click,” and “competition” columns and carries them into the results.

▶ Кластеризация семантического ядра: Бесплатный инструмент от SEOQUICK