Keyword research in 2026 looks nothing like it did five years ago. The keyword volume number in your favourite tool is more wrong than it has ever been. Top-of-funnel queries get answered by AI Overviews before users click. The “search intent” framework most blog posts teach was written for a search landscape that no longer exists.
This guide walks through the keyword research process we actually run for clients now. It works for content sites, ecommerce, and SaaS. It assumes nothing about your tool stack and explains why each step exists.
What changed
Three shifts make 2026 keyword research different from the textbook approach.
Volume numbers lie more than ever. AI Overviews siphon clicks from informational queries. The “10,000 monthly searches” number in your tool does not translate to clicks if AI Overview answers the query directly. Volume is still useful as a rough demand signal, not as a click forecast.
Entity coverage matters more than keyword targeting. Google ranks pages that comprehensively cover an entity well, even if the page does not exact-match the search phrase. A page about “internal linking” that explains the concept thoroughly will rank for hundreds of long-tail variations without specifically targeting them.
Zero-click queries dominate informational categories. If your goal is brand awareness via SEO, that is fine. If your goal is leads or sales, you need to focus on commercial-intent queries earlier in the keyword research process.
The process
Here are the eight steps we run for every keyword research project. Allow 4-8 hours per topic for the first time you do this.
Step 1: Define the business outcome
Before opening any tool, write down the single sentence answer to: “If this content piece works, what changes for the business?”
Not traffic. Specific outcome. Examples:
- “We get 30 free trial signups per month from this page”
- “We get 10 sales calls per month from this page”
- “We rank for terms our top competitor ranks for, taking their share”
This determines whether you should pursue informational, commercial, or transactional keywords. If you skip this step, you will produce a keyword list that ranks but does not move business metrics.
Step 2: List your seed terms
Open a blank document. List the words that:
- Describe your product or service in plain language
- Describe the problem your customers have before they need you
- Are the most common terms used in your industry
If you sell project management software, your seeds include “project management software”, “task management”, “team collaboration”, “agile project tools”, “kanban boards”. Aim for 15-30 seed terms.
Step 3: Expand seeds in a keyword tool
Plug each seed into a keyword tool (Ahrefs, Semrush, KWFinder, Moz, Google Keyword Planner). Pull the related keywords, questions, and “Also rank for” data.
For each seed, save:
- All matching keywords above 100 monthly searches
- All question-format keywords (these are valuable for AI Overview targeting)
- All long-tail variants that include modifiers (best, vs, alternatives, free, how to, examples)
You will end up with a list of 500-2000 keywords. Do not panic. The next steps cut this down.
Step 4: Cluster by intent and topic
Group your keywords into topic clusters, then within each cluster, group by search intent.
Intent categories we use:
- Informational: user wants to learn (“what is keyword research”)
- Commercial investigation: user is comparing (“ahrefs vs semrush”)
- Transactional: user is ready to buy (“buy ahrefs subscription”)
- Navigational: user is looking for a specific brand (“ahrefs login”)
A single piece of content should target one cluster of closely-related intent. A “What is X” article and a “Best X tools” listicle are different pages. Trying to combine them produces a worse page that ranks for neither.
Step 5: Pull SERP data for the top keyword in each cluster
This is where most keyword research goes wrong. Volume and difficulty are not enough. The actual SERP tells you whether you can rank.
For each cluster, pick the highest-volume keyword and search it in an incognito window. Note:
Are AI Overviews showing? If yes, click-through rates are 30-60 percent lower than the volume suggests. Adjust your traffic forecast.
What types of pages rank? If the top 10 are all listicles, your in-depth single-product page will not rank for that query. If the top 10 are video tutorials, a text article will struggle. Match the format that ranks.
Who is ranking? If three of the top five are domains with DR 80+ (Wikipedia, Forbes, HubSpot, Backlinko), the keyword is probably not winnable for a new site. Pick a longer-tail variant instead.
What is the user actually getting? Read the top 3 results. If they all answer a question your seed keyword does not literally ask, that is the real intent. Build for that intent.
Step 6: Filter for opportunity
Apply these filters in this order:
1. Drop everything where the top 10 has 5+ DR 70+ domains unless your site is also DR 70+. You will not win. 2. Drop everything where AI Overviews show and the query is informational unless you specifically want brand awareness. The clicks are not there. 3. Drop everything below 200 monthly searches unless it has commercial intent. Long-tail informational rarely returns the effort. 4. Keep everything with commercial intent above 50 monthly searches even if difficulty looks high. A “ahrefs vs semrush” page that ranks 8th still gets meaningful traffic and converts well. 5. Keep questions even if volume is low. Questions feed AI Overviews and featured snippets. Aim to answer 10-30 questions per pillar topic.
You should now have 30-100 viable keyword targets per topic.
Step 7: Map keywords to content
Each viable target keyword maps to one piece of content. The content is either:
- Pillar page: 3,000-6,000 word comprehensive resource for the highest-volume head term
- Cluster post: 1,500-3,000 word focused post for a sub-topic
- Listicle: comparison or “best of” content for commercial-investigation queries
- Question post: targeted answer to a high-volume question
Avoid mapping multiple keywords to the same page unless they have nearly identical intent. “Best free keyword research tools” and “keyword research tools free” are the same intent. “Keyword research tools” and “how to use keyword research tools” are not.
Step 8: Sequence by priority
Not all viable keywords are equal priority. Sequence by:
1. Bottom of funnel first (transactional, comparison): these drive revenue fastest 2. Top-of-cluster pillars next: build topical authority signals 3. Supporting cluster posts: fill out the topic, supporting the pillars 4. Long-tail questions last: these stack over time
For most clients we publish 1 pillar page and 4-6 cluster posts per month, plus 2-4 question posts. That is roughly 8-12 pieces of content monthly.
Tools we actually use
You can do all of this with Ahrefs alone. Here is the stack we use because each tool has a unique advantage.
Ahrefs: SERP analysis, traffic estimates, link data Semrush: keyword discovery for informational queries, intent classification Google Search Console: which queries you already rank for (cheapest source of high-intent keywords) Google Keyword Planner: paid search volume sanity-check AlsoAsked: question discovery Frase or SurferSEO: cluster mapping and brief generation
If budget is tight, the minimum useful set is GSC plus Ahrefs Lite or Semrush starter.
Common mistakes
The seven mistakes we see most often, ordered by how much they hurt:
Targeting volume over intent. A 10,000-volume informational keyword that gets 4 percent CTR after AI Overviews ate the rest produces 400 clicks. A 600-volume commercial keyword that converts at 8 percent into your funnel produces actual revenue.
Treating “keyword difficulty” as the only filter. Difficulty scores are blunt instruments. SERP analysis catches issues difficulty misses.
One keyword per page is wrong, but cramming 10 unrelated keywords is also wrong. One intent cluster per page. Group related variations under that cluster.
Ignoring branded competitor terms. “[Competitor] alternatives” pages drive qualified traffic with high commercial intent. Most clients sleep on these.
Skipping GSC analysis. GSC tells you the queries you already rank for in positions 5-30. Improving those pages is the fastest growth lever and most teams ignore it.
Forgetting that ranking is the start, not the end. A keyword that ranks but the page is poorly structured for conversion is a wasted ranking. Plan the content’s CTA, internal linking, and lead capture before writing.
Doing the research once and never refreshing. Re-cluster and re-prioritise every quarter. Search intent shifts. Your competitive set shifts. The keywords that mattered six months ago are not always the right ones now.
TL;DR
Define the business outcome first. List seeds, expand in tools, cluster by intent. Pull real SERPs to validate, not just difficulty scores. Filter ruthlessly. Map keywords to content one cluster at a time. Sequence bottom-of-funnel first. Refresh quarterly. Volume is a soft signal, not a click forecast. Commercial intent matters more than total volume in 2026.