Most guides to the basics of keyword research teach you to open a tool, type a word, and sort by search volume. That’s not keyword research — that’s browsing. Real keyword research is an exercise in subtraction: you start with hundreds of terms people actually type, then systematically throw away the ones you can’t rank for, can’t monetize, or that don’t match what a searcher wants. What survives is a short list of pages worth building. This guide walks the full loop, including the parts beginner tutorials skip: why every metric on your screen is an estimate, how difficulty scores are actually computed, and a fifteen-minute worked example you can copy today.
Keyword Research Is Mostly About Saying No
A new site owner runs a search, sees a keyword with 40,000 monthly searches, and gets excited. A practitioner sees the same number and asks three questions: who else ranks for it, what does the searcher actually want, and can I realistically land on page one this year. Usually the answer to at least one is “no,” and the term gets cut. The core skill in the basics of keyword research isn’t finding big numbers — it’s filtering a large candidate list down to the handful of keywords where demand, winnability, and business value overlap. Everything below is a filter you apply in sequence.
Start With Demand You Can Name, Not a Tool
Before you touch software, write down 8 to 15 seed terms from your own head. Seeds are the root topics your business serves, and they come from five reliable places:
- Customer language — the exact words prospects use in sales calls and support tickets, which rarely match your internal jargon.
- Your navigation — the products, services, and categories you already sell.
- Problems you solve — the pain that sends someone searching in the first place (“won’t turn on,” “how much does it cost”).
- Competitor menus — the topics rivals organize their site around.
- Autocomplete and “People also ask” — type a seed into Google and read what it suggests; that’s demand, unfiltered.
This step matters because a tool only expands what you give it. Feed it a lazy seed and you get a lazy list. Feed it the phrase a frustrated customer actually typed, and the expansion surfaces long-tail terms with real buying intent that no volume sort would have floated to the top.
Expand Seeds Into a Candidate List
Now bring in a tool. The mechanism is simple: a keyword database contains billions of queries mined from clickstream data and search-engine suggestions, and expansion pulls every phrase that contains or relates to your seed. Feed in your 8 to 15 seeds, and you should walk away with 300 to 1,500 candidate keywords. Use a multi-seed explorer if you have one — running seeds one at a time wastes hours and misses cross-topic overlaps.
Don’t clean the list yet. The goal at this stage is coverage, not quality. You want the messy, oversized pile so the filters that follow have something to cut. SEO Rocket does this expansion against live Ahrefs index data and clusters the raw terms by parent topic automatically, so you inherit a structured pool instead of a 1,200-row spreadsheet you have to detangle by hand.
The Four Numbers — and Why Each One Lies a Little
Every candidate carries four metrics. Learning what each one really means — and where it deceives you — is the difference between the basics of keyword research and cargo-culting a spreadsheet.
- Search volume is an estimate, not a meter reading. It’s modeled from clickstream samples and averaged over 12 months, so a term with strong seasonality (think “tax software”) shows a misleading flat average. Treat volume as an order of magnitude — 200 vs 2,000 vs 20,000 — not a precise forecast.
- Keyword difficulty is mostly a proxy for one thing: how many unique domains link to the pages currently ranking. It does not measure content quality, topical authority, or intent match, which is why a “hard” keyword with weak, outdated top results can be easier to win than its score suggests.
- Cost-per-click tells you commercial value. A high CPC means advertisers pay to appear, which means the traffic converts. A $12 CPC keyword with 300 searches often beats a $0.10 keyword with 8,000.
- SERP features — AI Overviews, featured snippets, shopping packs, “People also ask” — tell you how many real clicks are left. A query where an AI Overview answers everything above the fold may show 20,000 searches and deliver almost no clicks to organic results.
Search Intent Is the Cheapest Filter You Have
Intent is the fastest way to delete half your list. Every query maps to one of four intents, and matching it is non-negotiable because Google ranks pages that satisfy the dominant intent, full stop.
- Informational — “how to prune roses.” The searcher wants an answer, not a product. A sales page will never rank here.
- Commercial — “best pruning shears.” They’re comparing before buying; this wants a comparison or review.
- Transactional — “buy pruning shears.” Ready to purchase; this wants a product or category page.
- Navigational — “Felco pruning shears” — they want a specific brand, and you usually can’t intercept it.
The reliable way to confirm intent isn’t to guess — it’s to search the term and look at what already ranks. If page one for your keyword is all listicles and the top result is a 3,000-word guide, Google has told you the intent is informational. Trying to rank a checkout page there is wasted effort no matter how good your product is.
Difficulty Is a Proxy, Not a Verdict
Beginners treat the difficulty score as a gate: green means go, red means stop. That’s backwards. Because difficulty mostly reflects backlinks to the current top pages, you should override the number by actually looking at the SERP. The real question is: can I beat the weakest page on page one — the tenth result — not the market leader?
Open the top ten. If several results are thin, undated, off-intent, or from forums and low-authority sites, the score is lying to you and the keyword is winnable. If the top ten are all comprehensive, recently updated pages from established brands, a low difficulty score won’t save you. This “beat the weakest page-one competitor” benchmark is exactly what we build into SEO Rocket’s competitor analysis, because “outrank the #1 result” is a fantasy bar for a new or mid-authority site, while “beat result #10” is a Tuesday.
A Worked Example: One Seed, Fifteen Minutes
Say you run a small SEO consultancy and your seed is “keyword research.” (Numbers below are illustrative — yours will differ.) Expansion returns roughly 600 candidates. You filter:
- Cut on intent first. “keyword research jobs” and “keyword research meaning” don’t serve a consultancy — gone. That removes maybe 40% of the list in two minutes.
- Cut on winnability. “keyword research tool” is dominated by billion-dollar SaaS brands with millions of backlinks. You can’t win it this year — set it aside.
- Keep the long tail. “basics of keyword research” and “how to do keyword research for a new website” show modest volume, mixed page-one results, and clear informational intent — a guide can rank. “keyword research for ecommerce” has commercial value and a beatable SERP. These survive.
You started with 600 terms and ended with maybe 20 you’d actually build pages for. That ratio — a couple dozen keepers from several hundred candidates — is normal and healthy. If you’re keeping most of your list, you’re not filtering hard enough.
From List to Map: One Primary Keyword Per Page
Your survivors aren’t a to-do list yet — they’re a clustering problem. Group keywords by the intent and answer they share. “how to do keyword research,” “basics of keyword research,” and “keyword research steps” are the same page; you don’t write three thin articles competing with each other (that’s keyword cannibalization, and it splits your ranking signals).
Assign each cluster one primary keyword — the highest-value term that captures the group’s intent — plus 3 to 8 secondaries you’ll weave in naturally. The primary drives your title, H1, and URL; the secondaries add semantic depth and catch long-tail variations. One page, one job, one dominant intent. This mapping is where a spreadsheet becomes a publishing plan.
The Caveats Nobody Puts in the Beginner Guide
Three honest truths that keep the basics of keyword research from becoming superstition:
- The metrics are directional, not gospel. Volume and difficulty are models built on sampled data. Two reputable tools will disagree by 2–3x on the same keyword. Use them to compare terms against each other, not to predict exact traffic.
- Zero-click search is real and growing. AI Overviews and featured snippets answer many queries without a click. A keyword’s raw volume increasingly overstates the organic opportunity — always check the live SERP for how much real estate is left for a blue link.
- Research is a hypothesis, not a result. You can’t fully validate a keyword until you publish and watch it move. The best-researched target sometimes stalls; a term you almost cut sometimes takes off. Ship, measure, adjust.
Measuring, and Redoing the Work
Keyword research isn’t a one-time setup — it’s a loop. After you publish, give a new page 6 to 12 weeks before judging it; rankings jitter daily, so track the trend, not a single day’s position. Watch top-100 snapshots rather than spot checks, and cross-reference against Google Search Console and GA4 as ground truth, since index-based rank estimates are directional. SEO Rocket handles this tail end — rank tracking, AI-visibility tracking for whether you show up in AI Overviews, and a client dashboard — so the loop closes instead of dying in a spreadsheet.
Redo the full research quarterly. Demand shifts, new competitors enter, and terms that were unwinnable last year may soften as page-one results age. The playbook behind it is proven across 1,000,000+ ranking pages, and none of it relied on finding one magic keyword — it relied on running this filter-and-map loop consistently, at volume, on real data.
Frequently Asked Questions
How many keywords should I target per page?
One primary keyword and 3 to 8 closely related secondaries. They should all share the same search intent and answer — if two keywords would need genuinely different pages to satisfy them, they belong on different pages, not crammed into one.
Is search volume or keyword difficulty more important?
Neither in isolation. Volume tells you how much demand exists; difficulty (and the live SERP) tells you whether you can capture any of it. A low-volume term you can actually rank for beats a high-volume term buried under established competitors. Always weigh both against commercial intent.
Can I do keyword research for free?
Partly. Google autocomplete, “People also ask,” and Search Console show real demand at no cost, and they’re a legitimate way to learn the basics of keyword research. But they don’t give you volume, difficulty, or competitor gaps at scale — that’s where an index-backed tool earns its keep, especially once you’re mapping dozens of pages.
How often should I redo keyword research?
Refresh your target list quarterly and re-check individual keywords whenever a page underperforms after 12 weeks. Search demand, SERP layouts, and competitor strength all change, so a keyword map is a living document, not a one-time deliverable.
