Keyword Research: How to Do It in Six Steps, With an Example

Keyword research is the work of finding the exact phrases your customers type into a search engine, then choosing the ones your site can realistically rank for. The method fits in one line: collect the real phrasing, size the demand, read the intent behind each phrase, judge the difficulty, and give each chosen phrase its own page. Done well, it tells you what to write before you write it. Done badly, it produces a long spreadsheet and no traffic.

How to do keyword research in six steps

SEO keyword research is not a hunt for the biggest number. It is a matching exercise between what people search and what your business can answer better than the pages already ranking. Google's own starter guide frames it from the reader's side: experts and newcomers use different words for the same thing, and anticipating those differences helps your pages perform ^[1]. The six steps below work for a five-page site and for a large blog alike. Only the volume of data changes.

1. List seed topics. Write five to ten plain phrases a customer would say out loud, not your internal jargon.
2. Expand them. Collect variations from Google's search box, the People also ask questions and the related searches at the bottom of the results page.
3. Size the demand. Attach a monthly search estimate to each phrase so you can compare them.
4. Read the intent. Search each phrase and look at what ranks: guides, product pages, local listings or videos.
5. Judge the difficulty. Ask whether your site can compete with the pages on page one today.
6. Map phrases to pages. Give each page one main phrase and the close variants it can answer in the same piece.

Autocomplete is a fair place to start step 2, because Google says its predictions reflect real searches, adjusted for language, location and your past searches ^[2]. Run those checks in a private window.

The order matters more than the tool. Most wasted effort comes from sizing before reading intent, or from writing before mapping. Skip step 6 and two of your own pages end up competing for the same phrase. Our free keyword research method walks through steps 2 and 3 with Google's free tools.

Search volume, intent and keyword difficulty, read correctly

Three numbers drive every keyword decision, and each one is easy to misread. Treat them as signals that point in a direction, not as measurements.

Search volume is an estimate

Every volume figure you see is modeled. Ahrefs, for example, says its volumes blend Google Keyword Planner, Google Trends and other third-party data, refined by its own models. Its own study found those volumes roughly accurate for 60% of keywords against Search Console, and Keyword Planner's for 45%. Use volume to rank phrases against each other, never as a traffic promise.

Intent decides the format

Search intent is the job the searcher wants done. A phrase like best running shoes wants a comparison; running shoes size 10 wants a product page; how to lace running shoes wants a short guide or a video. The results page shows you the answer. If the top ten are all product pages, a blog post will struggle, however well written. Match the format that already ranks, then beat it on depth or clarity.

Keyword difficulty is a score, not a law

Keyword difficulty estimates how hard it is to reach page one. Each tool calculates it differently. Ahrefs bases its score only on the referring domains of the top ten results and ignores on-page factors. Mangools uses its own score built on SERP data. Semrush offers a Personal Keyword Difficulty fitted to your own domain. So the same phrase can look easy in one tool and hard in another.

A worked example: a local bakery

Picture a bakery in Austin that wants more custom cake orders from search. The owner starts with three seeds: custom cakes, birthday cakes and wedding cakes. Autocomplete and People also ask turn those into a list of about forty phrases, from custom birthday cakes near me to how far in advance to order a wedding cake.

Next comes intent. Custom cakes Austin shows a map pack and bakery websites, so it belongs on the bakery's main order page. How far in advance to order a wedding cake shows guides and forum threads, so it becomes a short article. Wedding cake ideas shows image-heavy galleries from national sites. That phrase is popular but hard, and it fits the bakery poorly.

Then difficulty. The local phrases face other Austin bakeries, not national publishers. That is a fair fight. The wedding guide competes with forums and planning sites, but the question is specific, and a baker answers it with real lead times. That is an edge.

The finished plan has three pages, not forty, and wedding cake ideas waits until the site can compete. Each one targets a main phrase plus the variants it can answer naturally. Six months later, Search Console shows which of them actually earn impressions.

YouTube keyword research: same method, different results page

YouTube keyword research follows the same six steps, but the evidence comes from a different place. Phrasing comes from the YouTube search box, intent comes from the videos that rank, and difficulty comes from how established those channels are. A phrase can be busy on Google and quiet on YouTube, or the reverse.

A quick check helps before you film. Search the phrase on Google first. If Google already shows a video block near the top, the topic has proven video demand, and a good video can earn traffic from both places. If Google shows only text results, a written page may serve the same searcher better.

Tutorials, reviews, comparisons and how to fix problems tend to suit video, because people want to watch the steps. Definitions and quick facts rarely do. Some keyword tools also pull YouTube data directly; Ahrefs' free Keyword Generator, for one, covers Google, Bing, YouTube and Amazon.

Common keyword research mistakes, and when it gets hard

Most keyword research fails for the same handful of reasons. None of them is about the tool.

• Chasing the biggest volume. A broad head term with a heavy page one is out of reach for most new sites. Longer, specific phrases face weaker pages and usually match what the searcher wants.
• Ignoring intent. Writing a guide where Google ranks product pages, or the reverse.
• One page per variation. Separate pages for keyword research tool and keyword research tools split your own strength.
• Stuffing the phrase. Google says keyword stuffing is against its spam policies, and its language systems understand pages that never repeat the exact words ^[1].
• Never checking results. Research is a hypothesis until Search Console shows impressions.

The method stays simple. The data gets hard. Free tools cover a small site well, but three jobs push you past them. The first is competitor keywords: Google's tools never show what another site ranks for. The second is difficulty at scale, because scoring hundreds of phrases by hand takes days. The third is grouping large lists into topics and pages. That is the point where a keyword tool earns its place.

Tools that do keyword research for you

Paid keyword tools do not change the method. They compress steps 2 to 5 into one screen and add the competitor view that free tools lack. Semrush, for instance, can run the whole workflow from seed to grouped list, and Ahrefs adds Traffic Potential and its own difficulty score so you can filter the ideas by both. If you already have a site, start with your own data: the Search Console Performance report shows the queries you almost rank for.

For a side-by-side look at every option, from free tiers to full suites, see our guide to keyword research tools, sorted by the kind of site each one fits.

The bottom line

Keyword research is less about finding hidden phrases and more about honest matching. Collect what people actually type, read what Google already rewards for each phrase, and pick the ones where your page can be the best answer. One main phrase per page keeps your site focused. Check Search Console after a few months and adjust. The tools speed up the collecting and scoring; the judgment about intent and fit stays with you.