How to Do Keyword Research for SEO (Step-by-Step)
A practical keyword research process for SEO — seeds, intent, clustering, and prioritization — plus free tools to find SEO keywords that convert.
Keyword research finds the phrases your buyers type into Google so you can build pages that match intent — not guesswork. Head terms like “keyword research” itself are dominated by free tools and mega-brands; this guide shows the process agencies use in 2026, then points you to our free keyword research tool and the broader agency playbooks pillar.
If you only remember one rule: map demand to URLs, do not collect vanity volume spreadsheets. A phrase with 50 monthly searches that converts beats a 10,000-search head term you cannot win or monetize.
Step 1 — Collect seed topics
List products, services, problems, objections, and locations you sell. Include:
- Service names customers actually say (not only internal product jargon)
- Problems you solve (“slow website,” “no bookings in shoulder season”)
- Competitor brand + category hybrids when relevant
- Queries you already impress for in Google Search Console
Export GSC queries for the last 3–12 months. Impressions without clicks are gold: they show demand you are almost matching. Pair this with a skim of what is SEO if your team is new to intent vocabulary.
Step 2 — Expand ideas
Use autocomplete, People Also Ask, related searches, competitor money pages, and a keyword tool. Capture SEO keywords, questions, and modifiers (near me, cost, best, vs, for [industry]).
Expansion sources that still work in 2026
| Source | What you get | Caveat |
|---|---|---|
| Google Search Console | Real queries you already touch | Biased to current site coverage |
| Autocomplete / PAA | Questions and long-tails | No reliable volume alone |
| Keyword databases | Volume + difficulty estimates | Estimates drift; validate with GSC |
| Competitor page titles | Commercial gaps | Do not copy thin clones |
| Sales / support transcripts | Exact buyer language | Low volume, high conversion |
Our keyword research clustering tool helps group ideas without drowning in CSV noise. For a wider stack, see best SEO tools 2026.
Step 3 — Label search intent
Intent mistakes create the wrong page type — and rankings that never convert.
- Informational — learn (
what is,how to,guide) - Commercial — compare (
best,reviews,agency,vs) - Transactional — buy / hire (
services,pricing,book,near mewith clear purchase intent) - Navigational — find a brand (
AnotherSEOGuru login-style queries)
Local modifiers often cut across commercial and transactional intent. A query like “hotel SEO agency Greece” is commercial-investigational; “SEO audit pricing” leans transactional. Match page template to intent before you write a word.
Step 4 — Cluster and map to URLs
Group synonyms and near-duplicates onto one page. Give each URL one primary phrase and a short list of natural variants for headings and FAQs. Avoid thin duplicates for every minor spelling or word-order change.
Clustering rules of thumb
- Same intent + same funnel stage → one URL
- Different intent → different URL (even if keywords look similar)
- Location variants often deserve their own pages when you truly serve those markets
- Hub articles support commercial pages via internal linking
Document the map in a simple sheet: primary keyword, secondary variants, URL, intent, status (exists / rewrite / new).
Step 5 — Prioritize what to ship
Balance three scores: volume, keyword difficulty (or competitive reality), and revenue proximity. New sites should attack commercial long-tails before national head terms.
| Priority band | Typical profile | Action |
|---|---|---|
| P0 | Commercial, realistic difficulty, clear CTA | Ship or rewrite first |
| P1 | Supporting how-tos that feed P0 pages | Build as cluster content |
| P2 | High volume, low conversion, heavy competition | Defer or treat as brand content |
| P3 | Vanity / off-offer | Ignore |
Refresh priorities after every major product or location change. Cross-check with how to increase organic traffic so research turns into shipping cadence, not slide decks.
Turning a cluster into a brief
A usable brief answers five questions before anyone opens a doc: primary keyword, search intent, who the page is for, what proof or differentiators must appear, and which internal links should pass authority. Add 3–5 FAQ candidates from PAA or GSC. Specify the desired URL and whether this is a new page or a rewrite. Without that, writers pad word count instead of matching demand — and you recreate the thin content problem keyword research was supposed to prevent.
Multilingual and seasonal notes
Tourism and local businesses often need parallel EL/EN maps, not a single English sheet translated late. Seasonality matters: summer ferry and villa queries spike months before arrival; winter “cost” and “agency” queries may be steadier B2B demand. Revisit seed lists before peak season, not after you miss the booking window.
Free tools vs paid suites
Start with Search Console + our free clustering tool. Add Ahrefs/Semrush-class databases when you publish weekly and need competitor gaps at scale — not on day one of a brochure site. Roundup: best SEO tools 2026. Technical blockers that hide otherwise good keywords belong in a technical SEO pass.
Mistakes checklist
- Choosing head terms you cannot win in 12 months
- Creating one URL per synonym and cannibalizing yourself
- Ignoring Search Console in favor of third-party volume only
- Writing informational blogs for transactional queries
- Stuffing primary keywords into every H2 without answering the query
- Never reconnecting research to content calendars and owners
- Skipping Greek/English (or other) language variants when you sell in multiple markets
If your research lives in a forgotten spreadsheet, you do not have a keyword strategy — you have archive clutter.
FAQ
How many keywords should a page target?
One primary intent plus natural variants in headings and FAQs — not a stuffed list. If two phrases need different answers, they need different pages.
How often should you refresh keyword research?
Quarterly for competitive niches; after every major product, location, or seasonality change. Always reconcile with fresh GSC data.
Are long-tail keywords still worth it in 2026?
Yes — especially for new domains and local businesses. Long-tails convert earlier while you build authority for harder head terms.
Should AI write my keyword list?
AI can brainstorm modifiers; it cannot replace GSC reality or competitive judgment. Use models to expand, humans to prioritize.
Ready to Get Started?
Use SEO content services, review agency playbooks, and get started with a scoped content and SEO plan.