If you want to know how to do keyword research properly, the steps are not secret. Backlinko’s keyword research guide lays out a sequence most practitioners would recognise: brainstorm topics, validate them in a keyword tool, analyse volume, difficulty, cost per click and trends, check search intent, target long-tail keywords, and review Google Search Console for quick wins. This post walks through each step and argues for a different emphasis than most guides take. Volume is the least useful number on the sheet. Intent and winnability decide whether a keyword is worth a page.
How to do keyword research: the six steps at a glance
Here is the process in one table, with the question each step answers and the mistake that most often goes with it.
| Step | The question it answers | The usual mistake |
|---|---|---|
| 1. Brainstorm topics | What does my audience talk about? | Starting in a tool instead of where people talk |
| 2. Validate in a keyword tool | Do people search for it, and in what words? | Accepting the first phrasing you thought of |
| 3. Analyse the metrics | Volume, difficulty, CPC and trend | Sorting by volume and stopping there |
| 4. Check search intent | What kind of page does Google show? | Writing a guide where Google ranks products |
| 5. Target long-tail keywords | Where can I realistically rank? | Dismissing low-volume phrases |
| 6. Review Search Console | What am I already close to ranking for? | Never looking at your own data |
Step 1 and 2: brainstorm, then validate
Start where your audience speaks in its own words. Backlinko suggests Google’s related searches, Reddit, forums and YouTube autosuggest. These sources show the phrasing real people use, the problems they are stuck on and the comparisons they make, which a keyword tool will only partly reflect.
Then validate. Put each idea into a keyword tool and look at the variations it returns. Often the phrasing you started with is not the one people search; a nearby variant has clearer demand or a more specific intent. This is also how to do keyword research for free at a basic level: autocomplete, related searches and the free tiers of most tools give you enough to confirm that demand exists, even if the numbers are rough.
Step 3: read the metrics without over-trusting them
Backlinko defines search volume as “How many people search for this each month” and keyword difficulty as “How hard it is to rank. The lower the score, the easier it should be.” Both definitions are fair. Both numbers are estimates produced by each tool’s own model, and different tools disagree, sometimes widely.
That is why volume should not lead. A high-volume term often hides a mix of intents, a results page dominated by large brands, or demand concentrated in a few months of the year. Our post on search volume covers why the figure misleads, and the one on keyword difficulty explains what the score does and does not measure. Use cost per click as a rough signal of commercial value and trend data to spot seasonality or decline.
The practical rule: use the metrics to rule keywords out, not to rank them in. A difficulty score far above what your site has achieved before is a reason to look elsewhere. A volume of zero in one tool is not proof nobody searches.
Step 4: check search intent before anything else is final
Search intent is usually grouped into four types: informational, commercial, navigational and transactional. The type is not something you decide. Google shows it to you on the results page. Search the keyword and look at what ranks: guides, comparison pages, product pages or a single brand’s homepage.
If the results are product pages and you planned a blog post, the keyword is not yours with that page, however good the numbers look. That single check removes more bad targets than any metric. Our guide to search intent covers mixed results pages, which are common and need more judgement.
Step 5 and 6: long-tail terms and your own data
Long-tail keywords are longer, more specific phrases with lower individual volume. They are where newer and smaller sites can actually rank, because the intent is clearer and fewer strong pages target them precisely. A set of related long-tail terms often adds up to more useful traffic than the head term you could not reach. Our post on long-tail keywords covers how to group them.
How to do competitor keyword research fits here too. Put a competitor’s domain into a keyword tool, look at the terms it ranks for that you do not, and filter for the ones that match your audience and intent. A structured content gap analysis turns that list into a plan instead of a pile.
Finally, look at what you already rank for. Google Search Console’s performance report shows the queries your pages appear for, with impressions and average position. Queries with plenty of impressions where you sit just off the first page are often the quickest wins: a better title, a missing section or an internal link can be enough. Our guide to using Google Search Console shows where to find them.
Mapping keywords to pages, and the limits
Research is only useful once each keyword has a home. Map every target to one page, and give each page one primary intent. Where two keywords return largely the same results, they belong on the same page; where results differ, they need separate pages. That map is what stops two of your own pages competing for the same query.
How long does it take to do keyword research? It depends on scope. Validating a single article’s target can be quick. Building a map for a new site or a new section takes longer, mostly because checking intent on each results page cannot be automated well. Treat it as a recurring task rather than a one-off: refresh the map when your Search Console data shows new queries or when a section stops growing.
The limitation is the data itself. Every volume, difficulty score and CPC you use is a third-party estimate, and none of them can tell you how your particular site will perform against a particular results page. The process reduces bad bets. It does not remove them. Publish, watch what Search Console reports over the following months, and feed that back into the next round of research. Your own impressions and clicks are the only figures on the sheet that are measured rather than modelled, so give them more weight each time you repeat the process.
Frequently asked questions
What are the main steps of keyword research?
Brainstorm topics, validate them in a keyword tool, analyse volume, difficulty, CPC and trends, check search intent, target long-tail keywords and review Google Search Console.
Can you do keyword research for free?
Yes, at a basic level. Autocomplete, related searches, forums, free tool tiers and your own Search Console data are enough to confirm demand.
What are the four types of search intent?
Informational, commercial, navigational and transactional. Check the results page to see which one Google associates with a keyword.
Should I target the highest-volume keywords?
Not by default. Intent and how realistic it is to rank matter more. Long-tail terms are often a better starting point.
The takeaway Brainstorm where people talk, use the metrics to rule keywords out, let the results page decide intent, and give your own Search Console data the final say.
