Keyword difficulty (KD) is a score, usually from 0 to 100, that estimates how hard it would be to rank on page one for a search term. Every major SEO tool publishes its own version, and almost all of them calculate it primarily from the backlink profiles of the pages currently ranking.
It is a useful filter and a poor decision-maker. This page covers what goes into the number, what every vendor leaves out, and how to judge a keyword when the score and the reality disagree — which they do often enough to matter.
What the score is actually measuring
Despite the name, most difficulty scores are not measuring difficulty. They are measuring links to the current top results, and calling that difficulty.
The typical calculation:
- Take the pages currently ranking on page one for the term.
- Look at how many unique domains link to each of those pages.
- Weight by how strong those linking domains are.
- Normalise to a 0–100 scale.
That is close to the whole model. Ahrefs is explicit that its KD estimates the number of referring domains you would need to reach the top ten. Others fold in extra signals, but link data dominates every commercial implementation.
Two consequences follow immediately, and they explain most of the frustration people have with the metric.
A term can score low and be unwinnable. If the ranking pages have few links but Google has decided the query deserves official sources, or large retailers, or a video carousel, a low KD is simply wrong about your chances.
A term can score high and be very winnable. If the ranking pages are strong sites that happen to answer the question badly — a common situation for specific, long-tail commercial queries — a high score is measuring their link profiles, not the quality of their answers.
The scores are not comparable between tools
This trips up more keyword research than any other single thing.
| Tool | Scale | Headline input |
|---|---|---|
| Ahrefs | KD 0–100 | Referring domains to the top-ranking pages |
| Semrush | KD% 0–100 | Link data plus SERP features and competitive density |
| Moz | Difficulty 0–100 | Page Authority of ranking pages, plus a CTR model |
| Mangools | KD 0–100 | Link profile strength of the top results |
Different crawls, different indexes, different formulas. The same keyword routinely comes back at KD 18 in one tool and KD 34 in another, and neither is lying.
The rule is simple: scores are only meaningful relative to other scores from the same tool. Pick one tool, build your whole list in it, and compare within that list. A spreadsheet that mixes Ahrefs KD with Semrush KD% is a spreadsheet that will send you at the wrong keywords.
What no difficulty score includes
This is the part that decides whether keyword research works, and it is absent from every vendor's number.
Search intent. The score does not know whether the ranking results are product pages, tutorials, comparisons, or definitions. If every result is a how-to guide and you intend to publish a product page, your real difficulty is not 22 — it is effectively infinite until you change format. Intent mismatch is the single most common cause of a well-optimised page never ranking.
SERP features and who owns the click. An AI Overview, a featured snippet, four ads, a video carousel and a People Also Ask block can leave the first organic result well below the fold. Two terms at KD 15 can have completely different traffic outcomes depending on how much of the page is left for you.
Brand and entity preference. For some queries Google strongly prefers recognised brands or official sources. No amount of links or content quality moves an unknown site onto those results.
Topical authority. A site that has published forty connected articles on one subject has an advantage on the forty-first that no link-based score captures. This cuts both ways — it is why niche sites beat bigger generalists, and why your first article in an unfamiliar topic is harder than the score suggests.
Freshness requirements. Some queries are refreshed constantly by Google. Ranking there means committing to ongoing updates, not publishing once.
Commercial value. Difficulty says nothing about whether ranking is worth anything. A KD 5 term with no buying intent can be a worse investment than a KD 45 term that converts, and CPC is the fastest available proxy for that — advertisers do not bid meaningfully on terms that do not convert.
How to judge a keyword properly
Use the score as a first-pass filter, then do the part the tool cannot.
1. Filter on KD to build a shortlist. This is what the number is good for — reducing ten thousand keywords to fifty worth looking at. Set the ceiling relative to your own site's link profile, not an absolute rule.
2. Open the results page. Not the tool's preview. The live one, in a private window. Nothing else in keyword research returns as much per minute spent.
3. Read the intent. What format is ranking? Guides, product pages, comparisons, forum threads, videos? If you are not prepared to produce that format, stop here regardless of the score.
4. Look for a weak result. If one page on the first page is genuinely poor — thin, outdated, off-topic, or a forum thread — that is your opening. A single beatable result matters more than a flattering difficulty score.
5. Check how much click is left. Count what sits above the first organic result. An AI Overview plus four ads changes the value of the position substantially.
6. Check the link gap. Compare referring domains for the ranking pages against yours. This is the check KD was compressing, done directly and more usefully.
7. Check commercial value. CPC and the nature of the term. Ranking for something nobody buys is a hobby.
Steps 2 and 3 catch the majority of expensive mistakes, and neither requires a paid tool.
How to check keyword difficulty without paying for it
Every paid tool gates difficulty behind a subscription, and the free tiers are usually capped at a handful of lookups a day. You can assemble a serviceable version yourself, and because you are reading the SERP directly it is often better than the score.
Google Keyword Planner gives you volume bands and, critically, top-of-page bid ranges. It is free with any Google Ads account and no spend is required. The bid range is the single best free proxy for commercial value — advertisers do not pay $40 a click for terms that do not convert. Note that volumes are bucketed into wide ranges unless you are actively spending, so treat them as orientation only.
Google Search Console is the highest-value free source you already own, and it is not really a difficulty tool — it is better than one. Filter queries by average position 8 to 25 and you have a list of terms Google already considers you relevant for. No difficulty estimate is needed for a term you already rank for.
The SERP itself, read properly, tells you most of what a score would. Count how many of the first ten results are large recognised brands. Look for forum threads, thin listicles, or pages dated three years ago — each one is a gap. Check how much of the visible page is taken by ads, an AI Overview, and People Also Ask before the first organic link.
Free-tier lookups, used sparingly. Ahrefs Webmaster Tools gives verified site owners backlink and Search Console-linked data for their own property at no cost. Ubersuggest, Mangools and Semrush all allow a few queries a day. Save these for the shortlist rather than burning them on exploration.
A crude manual difficulty check that works surprisingly well: search the term, and for each of the top five results note whether it is (a) a recognised brand, (b) a page precisely about your term rather than mentioning it in passing, and (c) recently updated. Five out of five means move on. Two or fewer means there is room.
None of this needs a subscription, and steps two and three are more informative than any 0–100 number.
Calibrating difficulty to your own site
A score is only meaningful against your own link profile, so set your ceiling deliberately rather than copying a rule from a blog post.
| Your situation | Sensible KD ceiling | Reasoning |
|---|---|---|
| New site, under 10 referring domains | 0–15 | Anything link-contested is out of reach; win on specificity |
| Established, 50–200 referring domains | 15–30 | Room for moderately contested terms in your niche |
| Strong niche authority, 200+ domains | 30–50 | Topical authority substitutes for raw link volume |
| Recognised brand | 50+ | Brand preference and existing authority do the work |
Two adjustments worth applying on top.
Adjust down for a new topic. Your site's overall authority helps less than you expect on the first article in an unfamiliar subject area. Treat your effective ceiling as a tier lower until you have published several connected pieces.
Adjust up for specificity. A four- or five-word term with clear intent is frequently easier than its score implies, because difficulty models are less reliable at low volumes and the ranking pages are often targeting a broader term incidentally. This is where the gap between the score and reality most often works in your favour.
Striking distance beats difficulty
The best keyword opportunity for most sites is not on a difficulty list at all.
It is the set of terms you already rank for in positions 8 to 25 — far enough down to get almost no clicks, close enough that a genuinely better page moves you into range. Google has already decided your page is relevant. You are not fighting for a foothold; you are improving a position you hold.
These terms are in Search Console, filtered by average position. For a site with any history, that report is a better source of work than any keyword tool's difficulty filter, because the hardest part of ranking — being considered relevant at all — is already done.
The framing worth carrying: difficulty scores answer "how hard would it be to get in". Search Console answers "where am I already close". For an established site the second question is much cheaper to act on.
On a WordPress site, this is the part RankHive does for you. Every week it reads your Search Console data, finds the queries where a page already ranks on the lower half of page one or on page two, and drafts the change: a new title, a meta description, an internal link or a rewritten section, with the query, position and impressions next to it. Nothing goes live until you approve it.
Related reading
Difficulty scores are built largely from the same link data as Domain Rating, which is why the two move together and share the same blind spots. For the on-page checks worth running before you commit to a term, the technical SEO checklist covers the sequence.
Your job is to use the score as a filter, then spend your time on the terms Search Console says you are already close on. Work through them by hand with the steps above, or let RankHive find those pages each week and ship the fixes after you approve them.