How to Do Keyword Research Step by Step

To do keyword research step by step, start with a clear business and audience goal, gather the language people use, review what the searches actually need, and map the useful opportunities to pages. The result should be a small set of justified content decisions, not a spreadsheet full of phrases nobody knows how to use.

This guide walks through that process using a hypothetical online bookkeeping service for US freelancers. The examples are illustrative and include no invented search-volume figures. You can apply the same method to another service or educational website by changing the audience, offer, and page requirements.

Step 1: define the project before collecting keywords

Write down what the business offers, who it serves, and which outcomes the website should support. In the example, the service helps freelancers maintain bookkeeping records. It does not offer every financial service, and the research should not drift into topics requiring expertise or services the business cannot provide.

Specify the target market and language. A US-focused project should use relevant US terminology and data settings where the tool allows them. If the service is local, geography becomes a separate important consideration. This guide focuses on the general workflow rather than the additional decisions involved in local search.

Also record exclusions. If the business does not sell software, a broad software-shopping query may be a poor service-page target. Exclusions save time because you can reject unsuitable ideas before they become elaborate briefs or expensive drafts.

Step 2: inventory the pages you already have

List the important existing URLs with a short description of each page’s purpose. Include service pages, useful articles, and relevant category or resource pages. You do not need a complicated crawl for a small site; a carefully checked manual inventory can be enough to begin.

The inventory helps answer whether an opportunity requires a new page or an improvement to an existing one. If a service page already owns freelance bookkeeping, a closely related phrase may belong there. Creating another service page just to use different wording can introduce overlap without adding value.

Record obvious content limitations, but keep them separate from confirmed performance problems. “The page does not explain the onboarding process” is a visible observation. “This missing paragraph caused the traffic decline” would require additional evidence.

Step 3: collect seed topics from the business and its customers

Seed topics are starting subjects, not final keyword targets. For the example service, they might include freelance bookkeeping, organizing business receipts, and preparing records for an accountant. Use real customer questions, sales conversations, support requests, and the terminology people use when describing their needs.

Keep a source column beside each idea. A phrase from a customer email, a search report, and an AI-generated suggestion are different kinds of evidence. Recording the source helps you decide what still needs validation and prevents brainstorming ideas from being mistaken for proven demand.

Do not make the initial list enormous. A manageable set of relevant starting subjects is enough to test the method. You can expand later after you understand which sources produce useful ideas and which repeatedly lead outside the business’s scope.

Step 4: expand the ideas using appropriate sources

Review related search suggestions, relevant questions in search results, and the language used on representative pages. If you have authorized access to Search Console, existing queries can reveal subjects for which the site already receives visibility. Treat those queries as observations about your site, not a complete list of all market demand.

Keyword tools can help expand and organize ideas. Google describes Keyword Planner as an advertising research tool that provides keyword suggestions and estimates. Those outputs can inform SEO research, but ad competition and campaign forecasts are not the same as organic ranking difficulty or promised SEO traffic.

You may also use a tool such as Ahrefs or Semrush if it is available to you. Record the selected country, language where relevant, and date. Do not combine figures from different databases as though they were measured under identical conditions.

Step 5: remove ideas that do not fit the offer

Review the expanded list against the project definition. Remove unrelated meanings, audiences the business does not serve, and products it does not offer. A keyword can be popular and still be inappropriate for the page you are planning.

For the bookkeeping example, a query seeking a software login is unlikely to be an opportunity for an independent service article. A question about organizing freelance receipts may fit an educational resource. A query asking to hire a bookkeeper may fit the service page rather than the blog.

Keep uncertain ideas in a separate review group instead of forcing an immediate decision. State what you need to clarify, such as whether the business serves that customer type. This preserves potentially useful ideas without allowing them into the active publishing queue prematurely.

Step 6: identify the task behind each promising query

Write one sentence explaining what the searcher wants to accomplish. Avoid using the keyword itself as the explanation. For “what records should a freelancer keep for bookkeeping,” the task might be understanding which records to organize before working with a bookkeeper.

Then inspect the types of results appearing for the query. A recurring pattern of guides, service pages, comparisons, or tools can help you judge the expected experience. Do not infer intent from one word alone or assume a tool’s automatic label settles the question.

If the evidence is mixed, record the uncertainty and consider a narrower query. The goal is not to find a perfect label for every phrase. It is to understand the intended page well enough to decide whether your website can provide a useful answer.

Step 7: compare similar queries before creating separate pages

Group queries that appear to ask for the same result. For example, different ways of asking what a bookkeeping service includes may belong on one service page or one focused explanation. Changing word order does not necessarily create a new content opportunity.

Compare the actual task, expected page type, and useful supporting information. If two proposed pages would use essentially the same outline and examples, investigate whether they should share one owner. If one query concerns choosing a provider and another concerns preparing records, the distinction may support different pages.

This is the point where research becomes a site-planning decision. The output should show which phrases support the same page and why. Avoid creating a separate URL for every row in the keyword export.

Step 8: interpret volume and difficulty with context

Use estimated demand as one input, not the sole ranking of opportunities. A narrow query can be relevant even when a tool reports little activity, while a broad query may attract many people outside the intended audience. Understand the estimate’s limitations before rejecting or prioritizing a subject.

Third-party difficulty scores can help flag competitive situations, but they are not Google metrics or guarantees. Review the actual results and consider the quality of the competing pages, the page type, and what useful contribution you can make. A low score does not mean an irrelevant page will rank.

Also consider whether searchers need to click a website to complete the task. Some questions can be answered directly in the search experience. That does not automatically make the topic worthless, but it affects the kind of outcome you should expect from creating the resource.

Step 9: prioritize by relevance, usefulness, and feasibility

Choose a small number of opportunities that connect clearly with your audience and offer. Consider the business value, the quality of evidence you can provide, the work required, and whether an existing page can be improved. A practical priority list should explain the reason behind the order.

For the example business, clarifying a core service page may come before a broad glossary article. An educational checklist that helps suitable customers prepare could also have a clear purpose. Neither choice needs a claim that it will generate a particular number of leads.

Use a simple explanation rather than a complicated score you cannot defend. “High relevance, clear page fit, and verified service information available” is more useful than an unexplained number calculated from speculative inputs.

Step 10: map each opportunity to a page and action

For every approved topic, record the page that will own it and whether the action is to create, improve, merge, or investigate. Include the primary reader task and any closely related queries that the same page should address naturally. This creates a bridge between research and actual website work.

Keep proposed URLs separate from live URLs. A suggested address is not evidence that the page exists. If an existing page is the best owner, preserve its current URL unless a change has a clear reason and an appropriate implementation plan.

The example below shows how a small research output can lead to different actions. The phrases and decisions are illustrative; they are not a substitute for investigating your own market and site.

Step 11: turn the decision into a brief

A brief should state the reader, purpose, format, important questions, and scope boundaries. It should also name the original evidence or business information available for the writer. The keyword is a useful reference, but it is not a complete instruction for creating a valuable page.

For a service page, confirm what the business actually does before drafting detailed promises. For an educational article, identify the example or explanation that will make the resource helpful. If the topic involves regulated professional advice, use appropriate expertise and avoid extending beyond the business’s role.

Include a relevant next step. An article preparing someone to work with a bookkeeper might link to the service process. A beginner SEO guide might offer a coaching session for reviewing a practice task. The continuation should fit the reader’s stage rather than interrupt the answer with a generic pitch.

Step 12: review the decision after implementation

Once the page is published or improved, keep a record of the date and changes. Check that the intended content and links are present and the page works as expected. Search performance may take longer to provide useful evidence, so distinguish implementation verification from outcome evaluation.

Review the queries and pages receiving visibility when enough data is available. Ask whether the audience and task still match the original brief. Unexpected queries can reveal a new opportunity or a mismatch, but a single impression is not a reason to rebuild the page.

If the content does not attract relevant traffic, investigate the broader situation. Topic fit is only one possible factor. The guide on why content is not getting traffic can help you avoid treating every weak result as a keyword-research failure.

Can you do keyword research without paid tools?

Yes. Customer questions, public search results, your own site inventory, and authorized Search Console data can support useful early decisions. Paid tools can make expansion and comparison easier, but they do not replace understanding the business or judging the page’s purpose.

Be honest about what you have not measured. If you have no reliable volume estimate, do not invent one. You can still explain that a topic addresses a repeated customer question or an important service decision, while labeling search demand as uncertain.

A small, well-reasoned research sheet is more useful than a large export with no page mapping. Start with the resources available, then add a tool when you can name the limitation it will solve.

How many keywords should one page target?

Choose one clear primary purpose and include related language that helps explain it. There is no universal number of phrases every page must contain. A comprehensive answer can naturally address several closely related queries without being designed around a fixed keyword count.

The key question is whether the supporting terms belong to the same reader task. If they pull the page into unrelated subjects, move them out of the brief. If they clarify the same question, cover them where they fit naturally rather than repeating them mechanically.

For help reviewing the choices in your own research sheet, one-on-one SEO coaching with Marium can focus on intent, page mapping, and practical priorities. Bring a small set of candidates and the pages they might belong to so the session produces specific decisions.