Doing keyword research well means working through a repeatable sequence — starting from seed keywords, expanding them into real variations, checking intent against what’s already ranking, evaluating volume and difficulty, then clustering and mapping the results to actual pages — rather than opening a tool, grabbing whatever has the highest volume, and writing toward it. This article is part of the What Is Keyword Research? guide.
The Four Things Worth Checking for Every Keyword
Before walking through the steps, it helps to have a mental checklist for judging any individual keyword candidate:
- Intent — why is someone actually searching this?
- Topic fit — does it belong in a cluster with other keywords you’re already covering, or is it its own thing?
- Competition reality — given your site’s current authority, can you realistically compete for this?
- Business alignment — if you ranked for it tomorrow, would it actually matter to the business?
A keyword that fails any one of these is rarely worth the effort, no matter how good its search volume looks in isolation.
Step 1: Start With Seed Keywords
Begin with the small set of broad, core terms your content is actually built around — not the exact phrases you expect to rank for, but the general topics underneath them.
→ Full guide: Seed Keywords: Where Keyword Research Starts
Step 2: Expand Into Real Variations
From each seed, expand outward using a few different sources at once, since no single one captures everything:
- Autocomplete and “People Also Ask” surface how people actually phrase a search, including question-based long-tail variations.
- A keyword research tool takes a seed and returns related terms along with volume and difficulty estimates at scale.
- AI chat tools are genuinely useful for the brainstorming and clustering part of this step — prompted with a topic and audience, they can generate dozens of raw keyword ideas and group them by likely intent quickly. What they’re not reliable for is the actual numbers: an AI model has no live connection to real search data, so a volume or difficulty figure it gives you is a plausible-sounding guess, not a fact. Treat anything a chat tool tells you about search volume as a hypothesis to check in a real tool, not a number to plan around directly.
→ Full guide: Keyword Research Tools: Free and Paid Options
Step 3: Mine Your Own Search Console Data
This step gets skipped constantly, and it’s often the fastest source of genuinely useful keywords: Google Search Console’s performance report shows which queries are already generating impressions for your site, sorted by clicks. Queries with meaningful impressions but few or no clicks are worth real attention — Google already considers your content relevant enough to show for that search, which means a targeted improvement to the page (a clearer title, a more direct answer, better-matched content) can capture traffic you’re already halfway toward earning, without any new content required.
Step 4: Check Search Intent Directly, Not by Assumption
For each keyword candidate that’s survived this far, look at what’s actually ranking for it right now. If the top results are all in-depth guides, a short page won’t compete regardless of how well it’s optimized. If they’re all product or service pages, an informational article is fighting the wrong format entirely. This single check — actually looking at the current results before writing anything — prevents a large share of the wasted effort that comes from building good content aimed at the wrong intent.
Step 5: Evaluate Volume and Difficulty — Then Verify Both Manually
Pull search volume and difficulty estimates for your surviving candidates, but treat both as a first filter rather than a final answer.
For volume: check the trend over the past year, not just the current monthly estimate — a term that looks solid today but has been steadily declining, or that spikes only seasonally, behaves very differently than one with steady, consistent demand. It’s also worth knowing that the large majority of search terms get very little individual volume by most industry analyses — which is exactly why clustering related long-tail terms into a single page (Step 6) matters more than chasing a handful of high-volume terms alone.
For difficulty: scores vary meaningfully between tools, since each calculates it differently. Actually looking at who’s currently ranking — their content depth, backlink profile, and overall domain strength — gives a more honest read on whether ranking is realistic than a difficulty number by itself.
→ Full guide: Keyword Search Volume: What It Means and What It Doesn’t
→ Full guide: Keyword Difficulty Explained
Step 6: Cluster Related Terms and Map Them to Pages
Group closely related keywords that reflect the same underlying intent into a single, thorough page rather than splitting them across several thin ones that end up competing with each other. Then map each cluster deliberately to a specific page — existing or planned — so the site’s structure reflects an actual plan.
→ Full guide: Keyword Clustering: Grouping Keywords the Right Way
→ Full guide: Keyword Mapping: Assigning Keywords to Pages
Step 7: Revisit on a Regular Cycle
Keyword research isn’t a one-time project. Search behavior, competitors, and even your own site’s authority shift over time, so what was the right keyword to target six months ago may no longer be. A quarterly revisit — checking which targeted keywords are actually performing, which new opportunities have appeared, and which have quietly declined — keeps the research current instead of stale.
→ For how this fits into the bigger planning picture: SEO Strategy
A Worked Example
To make the sequence concrete, here’s how it plays out for a hypothetical site publishing content about home coffee brewing:
A seed keyword like “pour over coffee” expands into variations including “pour over coffee ratio,” “best pour over coffee maker,” and “how to make pour over coffee taste less bitter.” Checking Search Console shows the site already gets impressions for “pour over ratio” without much content addressing it directly — a real signal. Checking the current top results for “how to make pour over coffee taste less bitter” shows short, thin answers dominating, which suggests real room for a more thorough page. Volume and difficulty checks confirm the term isn’t seasonal and isn’t dominated by sites with overwhelming authority. Rather than writing three separate thin posts, all three terms — plus a few more long-tail variations around bitterness and ratio — get clustered into one comprehensive guide, mapped to a single planned page instead of split apart.
Common Mistakes
- Sorting by volume first instead of checking intent, and ending up with traffic that doesn’t convert
- Trusting an AI chatbot’s stated search volume without verifying it in an actual keyword tool
- Skipping Search Console entirely, missing keywords the site is already halfway ranking for
- Ignoring seasonality, targeting a term that looked strong in a single monthly snapshot but is actually in decline
- Treating keyword research as a one-time task rather than revisiting it on a regular cycle
Frequently Asked Questions
How many keywords should I research before starting to write?
There’s no fixed number — the right stopping point is when you have a coherent cluster of related terms that reflect one clear intent, not a scattered list of unrelated phrases.
Is it okay to use AI tools for keyword research?
Yes, for brainstorming and clustering ideas — they’re genuinely useful there. For actual volume and difficulty numbers, verify in a real keyword tool or Search Console rather than trusting a number an AI model states, since it has no live access to real search data.
What if a keyword has good volume but I can’t tell the intent from the results?
That usually means the intent itself is genuinely mixed or ambiguous for that term — often a sign to either narrow the keyword to something more specific, or to build content flexible enough to serve more than one intent at once.
The Bottom Line
Doing keyword research well is less about finding a secret list of high-volume terms and more about running a consistent process: seed, expand, mine your own data, check intent directly, evaluate honestly, cluster deliberately, and revisit regularly. Each step catches a different kind of mistake the others miss on their own.
