Keyword research is the process of finding the words and phrases your customers type into search engines, then deciding which ones are worth targeting. Done well, it tells you what to write, what to build, and what to leave alone. Done badly, it fills your site with content nobody who buys from you is actually searching for.
In 2026 the job has changed. Search volume matters less than it did, AI Overviews have absorbed the easy informational clicks, and the queries that still send valuable traffic are the ones tied to a real decision. Modern keyword research is as much about qualifying demand as discovering it — separating the searches worth a page from the ones that will never pay you back.
Why Keyword Research Still Decides Everything
Every other SEO decision flows from this one. Your site architecture, your internal linking, the pages you build, the content you commission — all of it is downstream of which keywords you chose to chase. Get the targeting wrong and you can execute flawlessly on everything else and still lose.
The scale of wasted effort is easy to underestimate. Ahrefs’ study of over a billion pages found that roughly 96% of them get no organic search traffic from Google at all. Most of that failure is not bad writing or weak links — it is content aimed at keywords the author never validated.
In my experience auditing content programmes, the single biggest source of waste is not poor execution. It is good writing pointed at keywords that no buyer ever searches. A polished 2,000-word guide on a zero-intent phrase is worth less than three sentences on a query someone types with a credit card in hand.
That is the mental shift for 2026: keyword research is a prioritisation exercise, not a discovery one. Finding keywords is trivial now — any tool spits out thousands. Deciding which handful deserve a page is the actual skill.
Start With Seed Keywords, Not Tools
The mistake most people make is opening a keyword tool first. Tools amplify your starting point; they do not create it. If your seed keywords are generic, your whole research is generic.
Better seed keywords come from the language your customers already use. Read your sales call notes, your support tickets, the questions in your inbox, and the subject lines of emails that convert. Those are the exact phrases real buyers use — before a tool ever rounds them into tidy head terms.
Then layer in three cheap sources that most people skip:
- Google Search Console. The queries you already get impressions for are demand you have partially earned. Pages sitting on positions 5 to 15 are the fastest wins available.
- The SERP itself. Autocomplete, People Also Ask, and Related Searches are Google telling you, for free, what else people want around a topic.
- Competitor content. The keywords your competitors rank for — and the ones they do not — reveal both the table stakes and the gaps.
Only once you have a strong seed list do the tools earn their place, expanding each seed into its long-tail variations and giving you the volume and difficulty numbers to sort them.
The Five-Step Keyword Research Framework
Here is the process I run for every content programme. It turns a messy list of thousands of phrases into a ranked plan you can actually execute.
- Build the seed list. Pull 15 to 30 core phrases from customer language, Search Console, and your existing top pages. These are your topic anchors.
- Expand each seed. Run every seed through a keyword tool to generate long-tail keywords, questions, and modifiers. Expect thousands of rows. Do not filter yet.
- Cluster by intent. Group the raw list into tight clusters where every keyword shares the same underlying job. Each cluster becomes one page, not one keyword.
- Score for priority. Rate every cluster on three axes — search intent value, business fit, and how achievable ranking is given your authority. This is where most of the list gets cut.
- Map to pages. Assign each surviving cluster to a new page or an existing one, name the primary keyword, and note the intent so the writer builds the right page.
The discipline is in step four. A cluster with 10,000 monthly searches but no commercial intent loses to a cluster with 200 searches from people ready to hire. Volume is an input, never the verdict.
How to Judge a Keyword: Volume, Difficulty, and Intent
Three signals decide whether a keyword is worth targeting. Read them together — no single one tells you enough.
Search volume is the crudest signal and the one people over-weight. It estimates how many times a keyword is searched per month, but it hides intent entirely. It also misleads on the long tail. An Ahrefs analysis of nearly 4 billion keywords found that about 95% of them get 10 or fewer searches a month — which means the bulk of your real traffic will come from thousands of tiny queries no volume tool even shows individually.
Keyword difficulty estimates how hard it is to rank on page one, usually based on the backlink strength of the current top results. Treat it as relative to your own authority. A difficulty of 35 is a wall for a new site and a warm-up for an established one. Always sanity-check the number by actually looking at who ranks — sometimes a “hard” keyword is held by weak pages you can beat with a better answer.
Search intent is the signal that decides everything, and the one tools are worst at. It is the job the searcher is trying to get done. The fastest way to read it is to search the keyword in an incognito window and study the top ten results as a group. If they are all guides, the intent is informational; if they are product and comparison pages, it is commercial or transactional. Matching that intent is non-negotiable — this is the same discipline covered in my guide to search intent optimization, and it is where most keyword targeting quietly fails.
A simple way to combine the three: priority = intent value × business fit ÷ difficulty. A keyword only earns a page when the intent is valuable, it fits what you sell, and the difficulty is within reach. Miss any one and it drops off the list, however tempting the volume looks.
Keyword Clustering and Mapping to Pages
Once you have a scored list, the last job is deciding which keywords share a page and which need their own. This is keyword clustering, and getting it wrong causes two of the most common ranking problems I see.
Cluster too aggressively and you force conflicting intents onto one URL — a page trying to be a buying guide and a definition at once satisfies neither. Cluster too loosely and you split one intent across several thin pages, which triggers keyword cannibalization where your own pages compete against each other for the same query.
The clean rule is one page per intent. Group every keyword whose searchers want the same thing onto a single page, then pick the highest-volume variant as the primary keyword and let the rest support it naturally in the copy. A well-built page ranks for its whole cluster, not just the head term.
Mapping is where research becomes a plan. Every cluster gets assigned to a URL, tagged with its intent and primary keyword, and slotted into your content calendar. That map is also how you build topical authority — clusters that connect into coherent topics signal depth to Google far more than scattered one-off posts. This is the connective work a good content SEO programme lives or dies on.
Keyword Research for AI Search
AI Overviews and answer engines have not killed keyword research — they have sharpened it. The queries that used to send easy informational clicks now often resolve inside the AI answer, so ranking first no longer guarantees a visitor. That makes the scoring step matter more, not less.
The practical response is to weight your priority scoring toward queries where a click still has value: commercial comparisons, transactional searches, and problem-specific questions an AI summary cannot fully resolve. These are the keywords where users still leave the SERP to act.
At the same time, being cited by AI engines is now a goal in itself. To earn citations you still start from a query — you research the questions your audience asks an AI, then structure content with clear answers, definitions, and data an engine can lift. Keyword research is how you find those questions in the first place. If you want that side handled properly, it is worth reviewing your approach with an experienced freelance SEO consultant rather than guessing at what the engines reward.
The tools and the process are the same as they have always been. What changed is the verdict: in 2026, keyword research is less about which words have volume and more about which words still lead to a customer. Nail that judgment and everything downstream gets easier.
Frequently Asked Questions
What is keyword research in SEO?
Keyword research is the process of finding the words and phrases people type into search engines, then deciding which ones are worth targeting based on demand, difficulty, and business value. It tells you what pages to build and what to leave alone. In 2026 it leans less on raw search volume and more on qualifying demand — finding the queries where a click still leads to a customer rather than to an AI Overview.
How do I do keyword research for free?
Start with Google itself: type a seed keyword and read the autocomplete, the People Also Ask box, and the Related Searches at the bottom. Google Search Console shows the exact queries your site already gets impressions for, which is the highest-quality free source you have. Free tiers of Ahrefs Webmaster Tools and Google Keyword Planner add volume estimates. That combination builds a solid list without paying for a tool.
What is a good keyword difficulty to target?
It depends on your site’s authority. A new site should target difficulty scores under roughly 20 plus low-competition long-tail phrases, because it cannot yet compete on high-difficulty head terms. An established site with strong backlinks can chase scores of 40 and above. Difficulty is relative — always judge it against the actual pages ranking on page one, not the number in isolation.
How many keywords should one page target?
One primary keyword and a cluster of closely related variations that share the same search intent. A single well-written page naturally ranks for dozens or hundreds of long-tail variants of its topic, so you do not need a page per keyword. Build one page per intent — forcing multiple intents onto one URL, or splitting one intent across many URLs, both damage rankings.
Does keyword research still matter with AI search?
Yes, arguably more than before. AI Overviews and answer engines still need queries to answer, and they pull from pages that clearly match intent. Keyword research now doubles as a filter: it helps you prioritise commercial and transactional queries where a click still has value, and structure content so AI engines cite you. The process is unchanged — the scoring weights just shifted toward business value.
