Deciding which keywords are worth writing for

Keyword research produces a list. Keyword triage produces a decision. The gap between the two is where most content budgets are quietly wasted, because a list of 400 terms with volume and difficulty attached is not a plan, it is homework nobody has done yet.

Working notes on search workflow, published alongside the free tools at rankcrown.com.

These notes describe the triage I use to get from a research export to a short list of things actually worth writing. It is opinionated, and the main opinion is this: a difficulty score is an input, not an answer, and treating it as an answer is the single most common way to waste six months.

What a difficulty score is actually measuring

Every keyword difficulty metric in every tool is a model. It looks at the pages currently ranking for a query, summarises how strong their link profiles are, and compresses that into a number on a fixed scale. The details differ between vendors, which is why the same keyword can score 22 in one tool and 41 in another. Neither number is wrong. They are answering slightly different questions.

What matters is understanding what the model can and cannot see. It can see how many domains link to the ranking pages. It cannot see whether those pages satisfy the query, whether the intent is commercial or informational, whether the result set is dominated by forums, or whether a brand has an unassailable entity advantage on that term.

So the score is a useful first filter and a terrible final filter. Running candidates through a free keyword difficulty checker is how I cut a 400-row export down to something a human can read. It is not how I decide what to write.

The three bands I actually use

Notice that all three rows end with a manual step. The score sorts the pile. It does not pick from it.

The SERP tells you what the score cannot

Before writing anything, open the results page for the query and read it as a document. You are looking for four things, and none of them appear in any metric.

Intent. What kind of page is winning? If the top ten are all product pages and you plan to write a guide, you are not competing, you are submitting the wrong format. The reverse is equally true: a commercial landing page will not displace a results set made entirely of tutorials.

Result type saturation. If half the page is occupied by a video carousel, a shopping unit, and a set of forum threads, the number of positions available for a conventional article is much smaller than ten. A difficulty score of 15 means very little when there are effectively three organic slots above the fold.

Freshness sensitivity. If every ranking page was published or updated in the last ninety days, the query rewards recency, which means the page you publish will need maintenance forever. That is a real cost and should be priced into the decision.

Brand gravity. Some queries are effectively navigational even when they look generic. If the results are dominated by three well-known brands and a comparison page about those three brands, a new site is not entering a ranking contest, it is asking to be added to a category that already closed.

The score tells you how hard the door is to push. The results page tells you whether there is a room on the other side.

Check where you already stand

A step people skip: before treating a keyword as new, find out whether the site already ranks for it somewhere on page two or three. This changes the economics completely. Moving an existing page from position 14 to position 6 is usually cheaper than creating a new page and starting at nothing.

Spot-checking a handful of candidate queries with a SERP position checker takes a few minutes and regularly reveals that a third of the list is already partially covered. Those become optimisation tasks rather than writing tasks, which is a much better use of the same hours.

Be careful how you read a single position reading. Results vary by location, by device, and by personalisation, and they move day to day. One check tells you roughly which page of results you are on. It does not tell you your exact position, and you should not build a report on the assumption that it does.

The cannibalisation question

When a site already has coverage, the next question is whether adding another page helps or hurts. Two pages targeting the same intent do not double your chances. They split internal signals, confuse the engine about which page to show, and often produce a worse result than either page alone.

A quick way to detect this: search for the query with a site restriction and see how many of your own URLs the engine considers relevant. If three of your pages surface and they are all trying to answer the same question, you have a consolidation project, not a content gap.

Consolidation usually means picking the strongest URL, merging the useful parts of the others into it, and redirecting. It is unglamorous work that nobody wants to schedule, and it frequently outperforms a month of new publishing.

Whether you can support the page after publishing

A page does not rank because it exists. It ranks because it is reachable, relevant, and supported. The support part is where most small sites run out of road, and it is worth checking honestly before committing.

Two questions matter. First, can the new page be linked from somewhere that already has authority on your own site? A page with no internal links pointing at it is an orphan, and orphans are discovered late and treated as peripheral. Second, is there any realistic path to external references? Not a link building campaign necessarily, but some reason a third party would ever cite the page.

Looking at what already points to the site gives a rough sense of what is achievable. A backlink checker is enough for this stage: you are not doing a full audit, you are asking whether the domain has any external support at all, and what kind of pages have historically attracted it. If the answer is that nothing external points anywhere except the homepage, plan for informational content that can earn references, not commercial pages that cannot.

The technical preconditions nobody checks

There is a specific kind of failure where the keyword choice was correct, the article was good, and nothing happened because the page was never eligible. It is worth ruling out before you write rather than after.

Three checks cover most of it. Confirm the section of the site you are publishing into is not disallowed, which a robots.txt tester settles in one query. Confirm new URLs actually reach the sitemap, which a sitemap validator will show. And confirm the template emits a correct canonical and a unique title, which a meta tag analyzer reads in a single request.

These take about five minutes combined and they are the difference between a slow start and no start at all. All of the free utilities referenced here are collected at rankcrown.com/en/free-tools.

A short list beats a long one

The output of this triage should be uncomfortably small. From a four hundred row export I usually end up with somewhere between six and twelve terms worth acting on, and two or three of those turn out to be optimisation jobs on pages that already exist.

That feels like a poor return on a research session, and it is exactly right. The cost of writing an article is not the writing. It is the opportunity cost of the article you did not write, plus the ongoing maintenance of a page that will never earn its keep. A shorter list with clearer reasoning behind each entry produces better outcomes than a spreadsheet nobody opens twice.

Reviewing the decision later

Write down why you chose each term. Not the score, the reason: what the results page looked like, what format you decided to use, and what you expected to happen. Three months later, that note is the only way to tell the difference between a bad keyword choice and good choice executed badly. Without it, every disappointing page looks the same and you learn nothing from any of them.