Step 1: Start from seed terms, not the tool

It starts before any tool opens: with a plain-language list of what the business actually sells and the words a real customer would use to ask for it. Seed terms are the handful of core phrases everything else in the process expands from.

Write down, in plain English, every product, service, and problem the business solves. Not marketing language, the words a customer would actually use. A gym doesn't sell "fitness transformation solutions," it sells personal training, weight loss coaching, and a place to work out near where someone lives. Those plain phrases are the seed terms.

Google's own Search Central documentation makes the same point from the search side: it advises anticipating that "users who know a lot about the topic might use different keywords in their search queries than someone who is new to the topic," using the example of someone searching "charcuterie" versus someone searching "cheese board" for the same underlying need. A seed-term list should capture both the expert phrasing and the plain-language version, because a keyword tool can only expand what it's given. Feed it five vague ideas and it returns five hundred vague variations. Feed it fifteen specific, customer-language seed terms and the expansion is worth reading.

A realistic seed list for most small businesses runs ten to twenty phrases: the core service names, the problems those services solve, and the questions a first-time buyer would ask before purchasing. That's the input for step two.

Step 2: Sort every keyword by search intent

Every keyword implies a different kind of searcher: someone learning, someone comparing, or someone ready to buy. Sorting a keyword list by intent before anything else stops a page from being built for the wrong kind of visitor.

Once the seed list is expanded into a full keyword list (using a tool like Google Keyword Planner, Ahrefs, Semrush, or Ubersuggest to pull related terms, questions, and variations), the next job is sorting every keyword into one of four intent buckets: informational (someone learning, "how does X work"), commercial (someone comparing options, "best X for Y"), transactional (someone ready to buy, "X pricing" or "book X"), and navigational (someone looking for a specific brand or page).

Intent decides the content format, not just the topic. A commercial-intent keyword needs a comparison or a landing page with a clear next step. An informational keyword needs an article that actually teaches something, padding it with a sales pitch just weakens it for both the reader and the ranking. Mixing intents on one page is one of the most common reasons a page ranks for a keyword but never converts: it satisfies the search, then loses the visitor because the format doesn't match what they came for.

This is also where SEO and AI visibility start to diverge slightly. A classic search results page can show ten blue links across mixed intents; an AI-generated answer picks one framing and runs with it, so getting the intent right matters even more when the goal includes getting quoted directly, not just ranked (the full breakdown of which one your budget should fund first is worth reading if that tradeoff is live for your budget right now).

Step 3: Weigh search volume against keyword difficulty

Search volume shows how many people search a term monthly; keyword difficulty estimates how hard it is to rank for it, usually from how strong the current top-10 pages are. The best targets balance both, not just the biggest number.

Search volume is the average number of monthly searches for a term. It's useful, but it isn't the whole picture, and it doesn't even mean what most people assume. According to Ahrefs' keyword research guide, even a page that reaches the number one ranking spot "will rarely exceed 30% of its search volume" in actual visits, so a 10,000-a-month keyword is not a promise of 10,000 monthly visitors even at the top of the results.

Keyword difficulty estimates how hard a term is to rank for. Tool methodologies vary: Ahrefs' KD score, for example, is built purely from the number of referring domains linking to the pages currently ranking in the top ten and explicitly excludes on-page factors, while other tools may weigh in signals like content depth or domain authority. A brand-new site chasing a keyword with an entrenched, heavily-linked top ten is picking a fight it's unlikely to win soon, no matter how good the content is.

The practical move is to plot every keyword on two axes, volume and difficulty, and prioritize the ones that balance both: meaningful search volume, a difficulty score the site can realistically compete for given its current authority. A keyword with high volume and low difficulty is genuinely rare. Most real opportunities trade some volume for a winnable difficulty score, especially for a newer or smaller site. This is the same audit a paid SEO retainer runs every month against a client's actual authority level, not a generic scoring cutoff applied to every site regardless of size.

Step 4: Cluster keywords into topics, not a spreadsheet

Related keywords that pull near-identical search results are the same topic in Google's eyes, not separate pages. Grouping them into clusters turns a messy list of hundreds of terms into a handful of real content decisions.

A finished keyword list often runs into the hundreds of terms. Building one page per keyword isn't just impractical, it actively works against the site: near-duplicate pages competing for near-identical searches split authority instead of building it.

The fix is clustering: grouping keywords that Google already treats as the same topic. The signal to check is the search results page itself. If two keywords return nearly the same set of ranking pages, Google has already decided they're the same underlying question, and one page can answer both. Ahrefs describes this directly in its keyword research guide, comparing the search results for "whipped coffee" against "whipped coffee recipe" and finding the top-ranking pages nearly identical, evidence that the second term is a subtopic of the first rather than a separate page's worth of content.

Run this check across the full keyword list and it collapses fast: a long list of individual terms typically collapses into a much smaller set of real content topics, though the exact ratio varies by niche and seed list. Each cluster gets one primary keyword (usually the highest-volume, clearest-intent term in the group) and a set of secondary terms the page should also cover naturally in its headings and body copy.

Step 5: Map every cluster to one page

Each cluster gets exactly one destination: an existing page it should be added to, or a new page that doesn't exist yet. A cluster with no page mapped to it is research that never becomes traffic.

This is the step most keyword research skips, and it's the one that actually turns research into a content plan. For every cluster from step four, decide: does an existing page already cover this topic and just needs strengthening, or does this need a new page built from scratch?

Check the current site against every cluster before creating anything new. A cluster that overlaps with an existing page (even loosely) is usually a case for expanding and re-optimizing that page rather than publishing a second, competing one, splitting authority across two pages targeting the same cluster undoes the clustering work from step four. A cluster with no matching page at all is a genuine content gap and a candidate for something new.

A keyword list that never gets mapped to specific pages is just a spreadsheet. It doesn't rank anything, doesn't answer anything, and doesn't get cited by anything. The mapping step is what makes the previous four steps worth the time they took.

Step 6: Set the priority order, then revisit it

Not every mapped page gets built at once, so clusters get ranked by a mix of volume, difficulty, and business value. Search behavior shifts, so the list gets revisited on a schedule, not treated as a one-time job.

Once every cluster has a page mapped to it, order the list. Weigh search volume, keyword difficulty, and business value together, a lower-volume cluster tied directly to a paid service is often worth more than a high-volume cluster with only tangential relevance. Most teams can realistically ship a handful of new or refreshed pages a month, so the order matters as much as the list itself.

Keyword research also isn't a one-time project. Search behavior shifts as language, competitors, and even the products themselves change, and new AI-driven search surfaces are changing how some queries get typed in the first place. Revisiting the full list on a set schedule (quarterly is realistic for most small and mid-sized sites) catches new opportunities and flags clusters that have drifted or gotten more competitive since the last pass.

The part most keyword research skips

The six steps above sound simple written out, and the mechanics genuinely are. What's hard is doing all six in order, every time, instead of stopping after step two or three because a keyword list already feels like progress. A list isn't a plan until every cluster has a page, and a plan isn't finished until it gets revisited on a schedule.

Nine years. 600 plus clients. 50 plus still active. That track record exists because the unglamorous steps, the mapping, the prioritizing, the quarterly revisit, are the ones that turn a keyword spreadsheet into pages that actually rank and get read. If keyword research keeps stalling out at the spreadsheet stage, our SEO and AI visibility retainer runs this exact process end to end: research, content, schema, and a monthly report on what actually moved.