How to Choose Keywords for SEO: Filter, Score, Assign
Choosing keywords for SEO means running four filters in order: does this searcher intend to buy, does the term carry commercial value, can this site realistically compete, and is there exactly one page that should own it? Search volume comes fifth, not first. Score each candidate 0–3 on the three core dimensions, and assign anything above six to a named page before a word is written.
What is the best way to choose keywords?
The best way is to filter before you rank. Keep only queries where the intent matches your offer, where the term can produce a lead or a sale, and where your site can plausibly outrank the current results. Then score each candidate and assign it to a single page. Anything left without an owner is a distraction.
The picker panel
Treat these three settings as gates, not preferences. A candidate that fails a gate is not "low priority" — it is out, and the list gets shorter before it gets scored.
A query only belongs on your list if the page you would build can satisfy the next action. "Gasket material comparison" is a reading task; "custom gasket supplier" is a shortlist task. They are not the same page, and they should never share an H1.
Judgment beats a score here. Open the results and ask whether you could produce something better than the weakest page in the top five. If the answer is no, difficulty is not the issue — the offer is.
Estimate the value before the volume. One enquiry worth a five-figure order outranks a thousand curious readers, and it changes which pages you build first. Skip this step and the content plan becomes a traffic contest with no commercial ending.
intent fit (0–3)
+ commercial value (0–3)
+ feasibility (0–3)
= total / 9
Three points per gate. No weighting, no hidden multipliers, no tool-specific "keyword score" that nobody can explain in a meeting.
Twelve candidates, scored in one sitting
The sheet below took about forty minutes for an industrial supplier. Notice that the highest-volume term is not the winner, and that the head term is dropped rather than saved for later.
| Candidate | Intent | Value | Feasibility | Total | Verdict |
|---|---|---|---|---|---|
| custom silicone gaskets supplier | 3 | 3 | 2 | 8 | Build |
| food grade gasket sheet | 3 | 3 | 2 | 8 | Build |
| silicone gasket price per meter | 3 | 3 | 1 | 7 | Build with care |
| gasket material comparison | 2 | 1 | 3 | 6 | Park |
| rubber sheet | 1 | 0 | 1 | 2 | Drop |
Read the middle of the table, not the top. The parked term is the one that will become valuable after you have earned links on the two builders — which is a reason to keep the sheet, not to argue about it now.
One intent, one page
Every query in a cluster should point at the same page, and no second page should compete for it. When two pages target one intent, search engines must choose between them, and your links end up divided between two mediocre results instead of concentrated in one strong one.
Custom silicone gaskets
One promise, one page. Sizes, tolerances and the quote form live here; the standard explanation lives on the support page and links back.
1 intent · 1 page · 1 ownerWhere the candidates come from
Four sources, in this order. The first two are free and usually better than any purchased list, because they describe your site rather than the market in general.
Your own query report
Terms you already appear for, including the ones on page three. These are proven relevant — the only question is whether the page deserves the click.
Sales and support language
The phrases buyers use in emails and calls. They are rarely in a keyword tool, and they are usually the closest thing to a converting query.
Competitor page titles
Not to copy, but to see which intents your competitors thought worth a page. Gaps in their coverage are cheaper to win than their best pages.
A keyword tool, last
Use it to size and sanity-check, not to generate strategy. Every tool will hand you thousands of terms; none will tell you which page should exist.
Five ways the picker goes wrong
Sorting by volume first
The biggest number is usually the vaguest term, owned by an incumbent with ten years of links. Sorting by volume selects the hardest work, not the best work.
Treating related queries as synonyms
"Supplier" and "manufacturer" imply different checks to a buyer. If the results differ, the page should differ; if they do not, pick one and stop writing two.
Scoring without opening the results
A difficulty number is a model of the results page, not the page itself. Look at what actually ranks: forums, marketplaces and thin directories are openings, not walls.
Keeping every idea
A list of 400 candidates feels like progress and guarantees dilution. Twelve assigned terms beat four hundred maybes, because only assigned terms get a page.
Assigning two pages to one term
Usually a product page and a blog post, written months apart by different people. Merge, redirect or re-target — but decide, because the search engine will decide for you otherwise.
Bring us your candidate list
Send twenty terms you are considering and the pages that would own them. We will score them with you and flag the two that will waste the next quarter.
How the picker runs with a client
Research and assignment
Keyword Research Services covers the first half: query mining from your own reports, intent grouping, scoring and a named page per cluster. You get the sheet, the reasoning and the discarded list, because the reasons for dropping a term matter as much as the reasons for keeping one.
We deliver it as a working file, not a PDF. You should be able to re-score it next quarter without us.
Then the page has to match
A scored list is worthless if the page still reads like a brochure. On-Page SEO Services turns each assigned term into a title, a heading structure, an answer near the top and internal links that keep the cluster together.
We build the first one with you, then review the rest — the method has to survive without a babysitter.
Questions about choosing keywords
How many keywords should one page target?
One primary intent and any number of phrasings of it. If a secondary term needs different proof, a different format or a different next step, it needs its own page. Two unrelated promises on one URL is how pages end up ranking for neither.
Is low search volume always a bad sign?
No. A term with forty monthly searches and a five-figure order value can outperform a term with four thousand searches and no purchase intent. Score value before volume, then use volume only to choose the order of work.
Should I target long-tail terms first?
Usually yes, not because they are easier in principle, but because a specific page is easier to make genuinely better than a broad one. Start where you can win the argument with evidence, then climb.
What if two of my pages already rank for the same term?
Pick the page with the better intent match and the stronger links, then merge the useful parts of the other into it and redirect. Do not leave both live and hope one wins; the split is the problem.
Do I need a paid keyword tool?
Not to start. A query report, sales emails and a look at competitor titles will produce a usable list. Buy a tool when you need volume estimates at scale or a repeatable difficulty model — and treat its score as an input, never a decision.
How often should the list be re-scored?
Quarterly, and whenever a page launches. Scores move because the results page moves: a competitor may leave, a marketplace may enter, or your own page may finally have the links that make a hard term feasible.