How to Do Keyword Research for Topic Clusters in 2026

Most guides on how to do keyword research hand you a keyword list. That is the easy part. The hard part — and the reason so many content plans stall — is turning that list into a prioritized set of topic clusters you can actually publish against.

This guide walks through the full process: from seed collection, through expansion and scoring, into clustering, and finally into content briefs your writers can pick up without asking questions. The differentiator is a minimal data model and cluster-first prioritization we use to plan VarynForge's own marketing site.

What keyword research is, and why it matters in 2026

Keyword research is the practice of identifying the phrases real people type into search engines, then mapping those phrases to pages you can realistically rank for. It is a planning discipline, not a data-collection exercise. Done well, it tells you which topics deserve a dedicated page, which belong as a sub-section, and which are not worth touching yet.

The reason it still matters in 2026 is distribution. Most searches are rare: an Ahrefs analysis of roughly 4 billion keywords found that around 95% of keywords receive ten or fewer searches per month. That long-tail shape means head-term rankings alone will not move the needle — you need coverage across a full topic cluster before the compounding effect shows up in organic traffic.

Set goals before you touch a keyword tool

Open a keyword tool before you know the goal and you will export a spreadsheet you cannot prioritize. The first half hour of any research pass belongs to three questions.

First, whose problem are you solving? A single cluster usually maps to one audience segment — a role or a job to be done. Write it in one sentence before you start typing seeds.

Second, what is the business outcome for this cluster? Awareness clusters optimize for reach and brand surface area. Conversion clusters optimize for qualified pipeline. Mixing the two in one cluster is the fastest way to publish articles that neither rank nor convert.

Third, which search intents align to that outcome? Informational phrases feed awareness; commercial-investigation phrases feed conversion. Every keyword that lands in the final list has to map back to one of those two outcomes.

Tools and data inputs you'll need

No single tool covers the full process. Five categories are enough if you pick one from each:

• Keyword tools — Semrush, Ahrefs, or Google's own Keyword Planner for volume, difficulty, and variations. If budget is tight, start with a free tool rather than skipping this step.
• SERP analysis — open the top ten results for each target query and read the pages. The real intent signal lives in the content already ranking, not in the keyword metric.
• Competitor pages — pull the keywords two or three strong competitors already rank for. Most of your seed list is hiding inside their existing content.
• Analytics — Google Search Console reports what you are already ranking for and, crucially, what you are half-ranking for. Half-ranking queries in positions 11 through 30 are the highest-leverage targets you own.
• Internal search — if your site or product has a search box, its logs are primary keyword data no external tool can see.

How to do keyword research: a four-step overview

Every cluster follows the same four-step loop: seed, expand, cluster, prioritize. Do each step in order — skipping ahead creates work you will redo. The four steps below map to columns in the data model in the next section: treat this as the narrative, and treat the model as the schema.

Step 1: Gather seed keywords and hypotheses

Seeds come from three places: what your customers call the problem in their own words, what your product pages already say, and what your direct competitors rank for. Twenty to thirty seeds is plenty.

Pull three or four recent sales or support calls and copy the phrases buyers used verbatim — those phrases outperform anything invented at a whiteboard. Add the exact H1s of your competitors' highest-trafficked content. Then add three or four phrases your own team reaches for when describing the topic. You are looking for variety, not completeness. Variety feeds Step 2; completeness only matters after clustering.

Step 2: Expand keywords and capture metrics

Feed each seed into one keyword tool and let it generate variations. Pull the top fifty related suggestions per seed.

For each candidate, capture exactly four fields: monthly search volume, difficulty or competition score, dominant search intent (informational, commercial, transactional, navigational), and primary SERP features (featured snippets, video carousels, shopping units). Resist adding more columns now — anything beyond the four becomes noise you filter around. Google's Keyword Planner is sufficient for volume and competition; commercial tools layer difficulty and SERP features on top.

Step 3: Cluster keywords into topic clusters

Clustering is where most workflows break. A cluster is not a folder of synonyms — it is a group of queries that the top-ranking page on Google satisfies with a single result.

The fastest manual approach: sort the expanded list by the URL currently ranking #1 for each query. Keywords sharing the same #1 URL almost always belong in the same cluster. Commercial tools do this algorithmically by comparing SERP overlap. Aim for six to twelve clusters per topic. Within each cluster, mark the highest-volume query as the pillar intent and the rest as supporting angles for sub-sections or follow-up posts.

Step 4: Prioritize and map to content

Score each cluster on four axes, zero to three each: intent match with the business goal, estimated opportunity from summed volume, difficulty inverted so low competition wins, and business value of the downstream conversion path.

Sum the four into a twelve-point score and sort. Clusters above eight become pillar pages; clusters between five and eight become supporting posts or FAQ entries; below five stays on the backlog. For each prioritized cluster, write a one-line content brief: audience, intent, the answer the pillar promises, and three supporting queries the post must cover.

A minimal data model for cluster-first research

The entire process fits in a seven-column table. We use the same schema inside the VarynForge dogfood tenant that generated this article's plan:

1. Keyword — the raw query string exactly as searchers type it.
1. Cluster ID — the pillar topic this query belongs to.
1. Role — pillar, supporting, or satellite within the cluster.
1. Intent — informational, commercial, transactional, or navigational.
1. Volume — monthly search volume from your keyword tool.
1. Difficulty — the tool's 0-100 competition metric, normalized.
1. Priority Score — the twelve-point value from Step 4.

Anything else — competitor URLs, ranking history, brief outlines — belongs in a separate, linked table. Keeping the working table at seven columns is the difference between a plan your writers will follow and a spreadsheet no one opens twice. For the measurement columns that sit alongside this planning table, see our post on compounding vs. bleeding content.

Measure, iterate, and scale your clusters

A keyword plan is a hypothesis — the measurement cadence turns it into a strategy. Track three cluster-level metrics on a fixed 30-60-90 day cycle using Google Search Console.

• Day 30: impressions per cluster. Zero impressions means the pillar page has not been indexed or matches no live queries. Fix crawl first; retarget the cluster second.
• Day 60: average position for the primary query. A number between 11 and 30 means you are on page two — add internal links and strengthen on-page coverage. Above 30 means the cluster intent is wrong.
• Day 90: clicks and click-through rate. Positions 1-10 with low CTR point to a mismatched meta title, or a dominant SERP feature (featured snippet, video pack, shopping unit) you did not plan for.

Review quarterly. Retire clusters that stay below 1,000 combined impressions after six months. Double down on clusters already ranking positions 2-5 — one focused on-page refresh usually closes the remaining gap faster than any new-content investment.

Common mistakes and a publish-ready checklist

Three patterns show up in stalled keyword plans: chasing head terms before the long tail is covered, letting volume override intent, and writing one mega-post per cluster instead of a pillar plus supporting posts. The last pattern is the most expensive — it masks poor coverage behind one 5,000-word article that competes with itself for its own queries.

Before a brief leaves the research phase, every cluster should pass this checklist:

• The pillar query is documented with its intent and target SERP feature.
• At least three supporting queries share the pillar's top-ranking URL.
• Difficulty and opportunity scores are captured, not guessed.
• The cluster maps to a single business outcome, not both awareness and conversion.
• A specific tool stack — from our true-cost buyer's guide — is named for the final export.
• The brief states the one-sentence answer the pillar page commits to.

Frequently Asked Questions

What is keyword research and why is it important for content strategy?

Keyword research is the process of finding the phrases real searchers use and mapping those phrases to pages you can publish. It matters because it keeps content strategy tied to observed demand rather than to internal opinion or guesses about what audiences want.

How do I start keyword research for a new website or topic?

Start with twenty seed keywords drawn from customer calls, competitor H1s, and your own product pages. Run each through one keyword tool, capture volume and difficulty in a single table, then cluster the results before writing any briefs or drafts.

Which metrics should I capture when expanding keywords and how do I use them?

Capture four per keyword: monthly search volume, difficulty score, dominant search intent, and primary SERP features. Use volume for opportunity sizing, difficulty to gate effort, intent to match the cluster's business goal, and SERP features to plan the on-page format.

How do I group keywords into topic clusters and decide pillar vs supporting pages?

Group keywords that share the same #1 ranking URL in Google — those are queries Google already treats as one topic. Within each cluster, the highest-volume query becomes the pillar page; the remaining queries become supporting posts, sub-sections, or FAQ entries.

How should I prioritize keywords when intent, volume, and difficulty conflict?

Score each cluster on intent match, opportunity, difficulty, and business value, zero to three each, and sort by the twelve-point total. When two clusters tie, favour the one with the clearer intent — volume without intent alignment delivers traffic that does not convert.

How often should I revisit and update my keyword research and topic clusters?

Review cluster performance quarterly. Refresh the full keyword plan once a year, or sooner if a major SERP feature shifts in your industry. Retire clusters under 1,000 combined impressions after six months; refresh clusters stuck at positions 2-5 first.

Further Reading

• What is Keyword Research? A Complete Guide - Conductor
• Keyword Research: The Beginner's Guide by Ahrefs
• What a decade in SEO taught me about keyword research - HubSpot
• How to Do Keyword Research: A Beginner's Guide - Squarespace
• The Basics of Keyword Research - Iowa State CALS/LAS

Sources

• Ahrefs - Long-Tail Keywords: A Study of 4 Billion Search Queries
• Google Search Central - Creating Helpful, Reliable, People-First Content
• Google Ads - Keyword Planner
• Google Search Console - Overview

Key Takeaways

Keyword research in 2026 is a clustering discipline, not a list-making one. Collect twenty seeds, expand them with one keyword tool, capture the four metrics that matter, and cluster queries by the URL already ranking. Score each cluster on a twelve-point rubric, ship pillar pages for the top quarter, and measure at days thirty, sixty, and ninety. Keep the working table to seven columns and the plan stays something your team can actually execute — not a spreadsheet that rots the moment the research sprint ends.