The 6-step keyword research workflow
Here is how to do keyword research end to end, in one afternoon, without a $99/month subscription: start with seed keywords, expand them with free autocomplete data, pull rough volume for discovery, filter by search intent, cluster the survivors into topics, then prioritise by opportunity. The one step most beginners skip, and the one that separates amateurs from professionals, is manually reading the live search results before committing to any keyword. The SERP tells you the real difficulty and intent that no keyword-difficulty score can.
Work through the steps below in order. Each one narrows a long, messy list into a short list of keywords worth building pages for. For where this fits in the bigger picture, our full keyword research guide covers strategy end to end.
Step 1: Write down your seed keywords
Seed keywords are the 5 to 15 obvious words and phrases that describe what you sell or write about. They are the input for everything else, so they do not need to be clever. They need to be broad and accurate.
If you run a running-shoe store, your seeds are things like running shoes, trail shoes, marathon shoes, shoes for flat feet. Pull them from three places: how you describe your product, what customers type into your site search, and the labels your competitors use in their navigation. Do not filter yet. A bad seed costs you nothing; a missing one costs you a whole content area.
Step 2: Expand seeds into a raw keyword list (free)
Now turn each seed into dozens of real phrases people actually search. The free-first stack does this well:
- Google autocomplete — type your seed into the search box and note every suggestion. Add a letter (running shoes a, running shoes b) to force more variations. These are real queries ordered by how often people search them.
- People Also Ask — search your seed and expand the PAA boxes. Each question is a long-tail keyword with obvious intent, and clicking one loads more.
- Google Keyword Planner — the "Discover new keywords" tool returns hundreds of related terms from one seed, free with any Google Ads account. See the official Keyword Planner guide for setup.
Paste everything into a spreadsheet. It is normal to finish this step with 200 to 500 rows of duplicates and junk. That is fine; the next steps do the cutting. Autocomplete and PAA are also where most long-tail keywords hide, which is exactly what a new site should target first.
Step 3: Pull metrics, and know what Keyword Planner can and cannot tell you
You need a rough sense of demand so you do not write for terms nobody searches. Keyword Planner gives you this, but with a serious limitation you have to understand.
Without active ad spend in the account, Keyword Planner does not show exact monthly searches. It shows volume buckets, ranges like 1K–10K or 10K–100K. An analysis of 60 million keywords by Authoritas found Google reports these as roughly 60 fixed bucket ceilings, not precise numbers, and that a keyword often has to move 20 to 30 percent before it jumps to the next bucket. The practical problem: two keywords sitting in the same 1K–10K bucket could be 1,200 searches or 9,800 searches. You cannot tell them apart, so you cannot prioritise between them.
The fix is a rule: use Keyword Planner for discovery, never for prioritisation. It is excellent at surfacing terms you would never have guessed. It is useless for deciding which of two similar terms to write first. When you need the exact number to break a tie, that is where a cheap real tool earns its place. DeployFlare's free rank and keyword checker returns exact-volume figures, including for city-level and vernacular queries (Hindi, Tamil, Marathi) where global tools often report a false zero, starting free and scaling from ₹499/month rather than $99. Our keyword research feature does the exact-volume work Keyword Planner's buckets can't.
Step 4: Filter by search intent
Volume is worthless if the intent is wrong. A keyword like keyword research pulls people who want a how-to guide; a keyword like keyword research tool pulls people ready to sign up. If you sell software, the second converts and the first mostly does not.
Tag every surviving keyword with one intent label:
| Intent | What the searcher wants | Example |
|---|---|---|
| Informational | To learn something | how to do keyword research |
| Commercial | To compare options before buying | best keyword research tool |
| Transactional | To buy or sign up now | keyword research tool free trial |
| Navigational | A specific site or brand | ahrefs login |
Cut keywords whose intent does not match any page you are willing to build. If you are unsure what a term's intent actually is, the next step settles it. Our deeper guide on matching content to search intent covers the edge cases, and buyer-intent keywords are worth a separate pass if you sell something.
Step 5: Read the live SERP (the step nobody does)
Before you commit to any keyword, search it in an incognito window and read the first page. This single habit will change your results more than any tool. The keyword-difficulty number is a model; the SERP is the ground truth.
Look for three things:
- True intent. If your term is keyword research and page one is all step-by-step guides, Google has decided this is informational. A product page will not rank there no matter how good it is. The results tell you what format wins.
- True difficulty. Ignore the KD score for a moment. Who actually ranks? If it is ten domain-authority-90 giants with 4,000-word guides, that term is a two-year project. If you see forum threads, thin pages, or results that do not quite answer the query, there is a gap you can fill, a classic sign of a low-competition keyword.
- SERP features. Featured snippets, PAA blocks, and video carousels change how much traffic the blue links get. A term with a huge answer box may send fewer clicks than its volume suggests.
Google's own helpful content guidance says Search rewards content created to satisfy people, not to game rankings. Reading the SERP is how you see what "satisfies" means for that specific term. For the full method of reading difficulty from results rather than scores, see how keyword difficulty actually works.
Step 6: Cluster and prioritise
You do not want one page per keyword. You want one page per topic. Many keywords are the same search worded differently: how to do keyword research, keyword research step by step, and keyword research for beginners all deserve a single strong guide, not three thin ones.
Group your filtered keywords into clusters where every term could be answered by the same page. Pick the highest-volume term in each cluster as the primary keyword and treat the rest as secondary phrasings to work into headings and body copy. This is keyword clustering, and it is what lets one page rank for dozens of variations. Then map each cluster to a URL so you never write two pages competing for the same intent; that mapping step is keyword mapping.
Finally, prioritise. Rank your clusters by a simple mental formula: business value multiplied by realistic ability to rank (from your SERP read in step 5), divided by effort. Start with the clusters that are commercially useful and winnable now, not the highest-volume terms you have no chance against.
A quick example, start to finish
Say you sell budget wireless earbuds. Seed: wireless earbuds. Autocomplete and PAA expand it to wireless earbuds under 2000, best wireless earbuds for calls, earbuds for small ears, and 300 others. Keyword Planner confirms wireless earbuds under 2000 sits in a healthy bucket. Intent is transactional, a buyer with a budget. You SERP-check it: page one is thin listicles, no strong brand pages, a PAA block. That is a winnable, high-intent term. It clusters with under 3000 and cheapest wireless earbuds into one buying-guide page. It goes to the top of your list. That is the whole workflow in one keyword.
Where the free stack stops working
The free stack is genuinely enough to run this process once. It breaks down when you do it repeatedly. You cannot compare two keywords inside the same volume bucket, you cannot see exact difficulty across a whole list at once, and you cannot track whether your published pages actually moved. At that point a cheap real tool pays for itself: exact volumes to break ties, difficulty across your full list, and rank tracking to close the loop. For the wider method, including competitor keyword analysis and seasonal keyword research, start from the full keyword research guide. If you are weighing tools, our Ahrefs alternative comparison lays out the price difference honestly.
Keyword research is not a tool problem. It is a process problem. Run these six steps in order, read the SERP before you commit, and you will out-research most sites paying ten times what you do.