How to Do Keyword Research for SEO

Keyword research has a reputation for being a data problem. It isn't. It's a decision problem: which handful of phrases should this page be about, given what people type and what your site can realistically beat?

You can answer that with free data. Here is the process, in the order to run it, with the exact questions to ask at each step.

Step 1 — Start from problems, not phrases

Write down the ten questions customers ask before they buy, in their words. Listen for the phrasing you don't expect: "is it worth it", "how much does X cost", "what's the difference between", "do I need". Those are the seeds.

If you can't produce ten, you don't have a keyword problem — you have a customer-conversation problem, and no tool will fix it. Read your own support inbox, your competitors' reviews, and the questions in the community forums where your buyers hang out.

Step 2 — Mine autocomplete and the "people also ask" box

Autocomplete is a live sample of what people type. It costs nothing and it reflects real volume, because suggestions only surface for phrases with enough searches to be worth predicting.

The procedure:

1. Type your seed into Google and note every suggestion.
2. Add a letter or a question word to the front: how, what, why, can, does, best, vs.
3. Repeat for each suggestion that looks like a real question.
4. On a result page, open "people also ask" and expand two or three entries — each expansion reveals more questions.

Thirty minutes of this produces fifty to two hundred candidate phrases for almost any niche. Example, from a seed of "keyword research": how to do keyword research for seo, how to do keyword research for free, keyword research for beginners, how to find low competition keywords, keyword research without paid tools.

Step 3 — Check what already brings you impressions

If your site has been live for more than two months, Google Search Console is the highest-quality keyword data you will ever get, because it's about your site specifically. Open Performance → Queries, then filter to positions 8–30.

Those are phrases where you're already close, where search engines already think you're relevant. Improving the page that ranks at 15 is usually the cheapest traffic available to you — far cheaper than a new page starting from nothing. Before you research anything new, block out an hour to improve two pages sitting on page two.

Step 4 — Read the intent, every time

Search the phrase. Look at the first five results and answer three questions:

• What format are they? Guide, listicle, comparison, product page, calculator, forum thread, video, map results.
• What are they assuming the searcher already knows? That tells you how deep the page needs to go.
• What would make someone go back to search after reading one? That gap is your opening.

Then classify the phrase:

A mismatch here is the most common reason a well-written page never ranks. If the results are all comparison tables and you wrote an essay, you're answering a question nobody in that slot asked.

Step 5 — Score difficulty by hand

You don't need a difficulty score from a tool, and the free ones are unreliable for small sites anyway. Judge the actual competition:

• Are the top results big brands, or small independent sites? Small sites in the top five means the phrase is winnable.
• Is any result visibly thin — a forum thread, a 300-word page, an outdated article with dead links? That's a target.
• Do the results answer a slightly different question? Precision is an advantage.
• Is there local intent, or a niche modifier your site uniquely covers? Narrower phrases are both easier and better-converting.

A simple three-way rating is enough: Winnable (small sites, dated pages, thin answers), Possible (mixed results), Not yet (major brands and definitive resources). Spend 80% of your effort on Winnable, and revisit Not yet in a year.

Step 6 — Assign one phrase per page, and write

Now make assignments. Each page gets one primary phrase, the intent it matches, and three to five close variants that belong on the same page because they'd be answered by the same text.

Watch for two mistakes:

• One page per variant. "SEO for beginners", "beginner SEO guide", and "SEO basics for beginners" are one page, not three. Splitting them creates pages that compete with each other and none of them wins.
• One page for everything. "SEO" as a single page is a page about nothing in particular. Specific questions deserve their own pages.

Then order the assignments: cheapest win first. A page targeting a Winnable phrase with clear intent, where you have genuine expertise, is the right first piece. Publish it, wait three or four weeks, look at impressions, and let that data tell you which assignment comes second.

What search volume is actually for

Volume tells you how many people ask a question. It does not tell you whether you can answer it, whether they'll buy, or whether the competition is beatable. Beginners over-index on volume because it's the only number in the tool.

A practical use of volume: sorting. If two phrases are equally winnable and equally aligned with what you sell, pick the one with more searches. That's the whole decision. A phrase with 40 monthly searches and a clear buying intent will out-earn a phrase with 4,000 searches from people who want a definition and will leave.

A worked example

Say you run a small bookkeeping practice. Seeds: what clients ask you.

1. Seeds include "do I need a bookkeeper", "bookkeeping vs accounting", "how much does a bookkeeper cost".
2. Autocomplete adds "for small business", "in my area", "or accountant", "for freelancers", "monthly".
3. Search Console (if the site is live) shows impressions for "bookkeeper for sole traders" at position 18 — an immediate improvement target.
4. Intent check on "how much does a bookkeeper cost" shows pricing guides and a few local firms: comparison intent, format is a range with what changes it.
5. Difficulty: mixed results with small firms ranking, one dated article — Winnable.
6. Assignment: one page, "How much does a bookkeeper cost?", with a real price range, what moves the price, and what's included. Three variants folded in: cost per month, cost for a small business, average bookkeeping fees.

That's a complete research cycle, done in under an hour, using nothing that costs money.