Keyword Difficulty (KD)

Two vendors define it in different terms. Semrush describes its score as a measure of the SEO effort it would take for content to rank in the top ten results on Google, from 0 to 100, and calculates it from factors such as the median number of referring domains pointing to the ranking URLs, the median ratio of dofollow to nofollow links, the median authority score of the ranking domains and search-results-page qualities of the keyword itself^[1]. Ahrefs describes an estimate of how hard it would be to rank on the first page, also from 0 to 100, based on the number of referring domains of the top ten organic results, and says it does not take on-page factors into account^[2].

So the same keyword has different scores in different tools, and none of them is Google’s. Google’s own keyword tool is built for advertisers: Keyword Planner gives estimates of monthly searches and the cost of showing an ad for a keyword, for Search campaigns^[3]. A difficulty score is a vendor’s estimate, worked out from the vendor’s data.

A score is also a snapshot. Ahrefs calculates it from the pages it sees ranking when it last refreshed the results page, shows a cached score in grey once the page has not been updated for 20 days or more, and refreshes keywords with higher search volume more often^[2]. A figure for a keyword with little volume can be weeks old.

The vendors say how to use it. Ahrefs advises against basing all SEO decisions on its score alone, and calls it a first filter for removing keywords that are too competitive or a way to prioritize the rest^[2]. Semrush groups scores in six levels, from very easy to very hard, with the advice that harder levels need more links, as well as better content^[1].

Read what sits under a single number. In QueryRev’s study, the share of sites that block at least one AI crawler by name is 14.2% overall, 17.6% among places 1 to 1,000 of the panel and 13.9% among places 1,001 to 10,000^[4]. One score for a whole results page hides the same kind of split.

The score is read in a tool. What a team can measure is whether it was right: record the score, the tool and the date when a keyword is chosen, and the position the page reached after a fixed time, such as six months. A tool whose scores do not line up with the positions a site reaches is a poor guide for that site, whatever its method.

Compare tools on a sample. Put the same twenty keywords through two tools and list those on which they disagree by a wide margin. A disagreement is a reason to open the results page, not to average the two.

Check the results page itself, from the country and device that matter. Note who ranks and what kind of pages they are, whether they answer the intent of the query, and which weak results hold a place in the top ten. A weak page that ranks is the clearest sign that a keyword can be won.

Sample data: an invented site, to show the method.

A sample bed shop runs three keywords through two tools. Tool A scores “bed frame sizes” at 24 and Tool B at 58. The tools disagree by a wide margin on two other keywords, in opposite directions. The shop does not average the scores.

It opens the results page for “bed frame sizes”. The top ten holds two forum threads and an article from 2019, none of which answers the question with a table of sizes. The shop judges the keyword winnable on the evidence of those pages, not on either score, writes the guide, and records both scores, the date and its own verdict, so that it can read them against the position the page reaches.