Keyword Research the Right Way

Research done without business context is guesswork. Before any keyword work, collect the following.

From the website and your own research

• What is sold, and how the offers are grouped

• How the industry works and how buyers buy

• Who the competitors are and how they describe themselves

From stakeholders, especially sales

• Which products or services matter most to revenue

• Which customers you want more of, and which you don't want

• What customers call things before they learn your terms

• The questions prospects ask on first calls

• Where you operate, if location matters

• What is launching, seasonal or being retired

By the end you should be able to write a short description of how the industry works, how the business sells, how customers talk and how the business talks. The gap between those last two is where much of the opportunity sits.

Break the research into its parts before searching for anything.

A seed keyword is the core noun: the product or service in its most basic form. It is the lowest common denominator across a family of searches. In "personal training," "one on one personal training" and "affordable personal training," the seed is "personal training."

A modifier is a word that changes what the searcher wants. "Affordable" and "one on one" are modifiers. So are colors, sizes, materials, locations, audiences, use cases and words like "best," "near me" or "for beginners."

Make one list of seeds and modifiers for each area of the business you plan to research: each product line, service or audience.

Be exhaustive with synonyms: "sneakers," "trainers," "tennis shoes." Search engines do not always treat synonyms as equal. And reduce what you see to its parts. If a competitor ranks for "best yoga mats for beginners," that gives you the seed "yoga mat" and the modifier "beginner," and with them every search that combines the two. The lists keep growing as the research turns up more.

Work through one seed theme, finish it, then move to the next. Start broad, with the seed by itself. The results will be long and noisy, and that is useful: they show variations and modifiers you had not thought of. Then narrow with your modifiers and read every keyword instead of sorting by volume, keeping the ones that pass the tests in Step 4. Record what you keep, and note the patterns you see along the way. They feed back into the seeds and modifiers.

Reading keywords one at a time is slow. It is also where the understanding comes from. Bulk exports skip the part of the work that produces insight.

You are not trying to collect every keyword. You are collecting every keyword that could plausibly bring the business a customer. Apply these tests.

This step is market research. You are learning how the world searches. Whether the site has a page for a keyword today is a question for later.

Imagine you typed the keyword exactly as written. What would you be looking for? Is it reasonable to think that person could someday do business with this company? Be realistic, not hopeful.

Some keywords carry clear intent: "find a skincare specialist near me." Others are too vague to read: "skincare." Is that a product, an article, a doctor? Vague, short keywords also tend to be dominated by sources that are not businesses. They are usually best left out.

A raw list cannot be analyzed. Tag each keyword so the landscape can be sliced.

The tags that matter most:

• Category, matching how the business groups its products or services

• Location, for businesses with physical locations or service areas

• Intent type. Transactional means the person is clearly looking for a product or service. Educational means they are looking for information.

Beyond those, tag the patterns specific to the industry, such as size, material, audience or occasion. Tags exist to reveal patterns. Add one only if it helps you or a stakeholder see something.

With tags in place you can answer questions like "what are all the valuable keywords about sneakers?" or "where is demand strongest in the Northeast?" They also make the landscape presentable to people who will never read the full list.

This step turns research into strategy.

An intent group is a set of keywords that express the same underlying need. "New white tennis sneakers," "white sneakers for tennis" and "best white tennis sneakers" are three phrases and one need.

The hierarchy:

• Audience group: who is this person? Tennis players

• Intent group: what do they want? White tennis sneakers

• Keywords and prompts: how do they say it? "white sneakers for tennis"

Single keywords are too granular to plan around. Treating each as its own problem leads to duplicated effort and pages that compete with each other. Broad categories are too vague to tell you what to build. Intent groups sit where planning works: specific enough to map to one page, broad enough to cover every phrasing of the need.

Present the landscape to stakeholders before finalizing it. Show it by category and by intent group, not as a flat list.

Ask:

• What is missing?

• What is here that we would not want to be found for?

• Which groups matter most to the business this year?

• Do any of these terms mean something different in our industry?

Record what was added or removed after this round. The first qualification for any keyword is whether it could lead to a productive sales conversation, and the people in this meeting are the ones who know.

Translate intent groups into AI prompts. People ask AI assistants for the same things in longer, conversational language. For each intent group, write a small set of prompts that express the same need the way a person would ask it. More than one, because phrasing varies and answers vary.

Load everything into tracking. Keywords go into a rank tracker and prompts into an AI visibility tracker, both tracked against the same intent groups. Record competitors too.

That first reading is your baseline: where the business stands today in search and in AI answers. Everything you do later is measured against it.

With the landscape built and the baseline set, you can see:

• Intent groups with no page. These are often the fastest wins.

• Intent groups with a page that ranks poorly. The page may be off-topic, or relevant but short on authority.

• Intent groups where competitors appear and you do not.

• Where demand is concentrated, by category, location or audience.

Sometimes the landscape shows there is little search demand for what the business sells. That is a valid finding. It tells you to put the budget elsewhere.

• [ ] Stakeholder conversations done, including sales

• [ ] Seeds and modifiers listed for every research area, with synonyms

• [ ] Keywords collected one theme at a time

• [ ] Each keyword checked for intent, winnability and then volume

• [ ] Difficulty scores ignored

• [ ] Keywords tagged so the landscape can be sliced by category, location and intent type

• [ ] Keywords grouped into intent groups, each one need that one page can satisfy

• [ ] Landscape reviewed with stakeholders and revised

• [ ] A small set of AI prompts written for each intent group

• [ ] Keywords and prompts loaded into tracking, with competitors

• [ ] Baseline recorded