Keyword Difficulty

Keyword difficulty (KD) is a score, typically 0 to 100, that estimates how hard it would be to rank for a given query. Every major SEO tool has its own version. Used sensibly, KD accelerates keyword prioritisation. Used as a single deciding metric, it produces bad strategy.

What does KD measure?

Most KD scores are calculated from the link profile of the pages currently ranking for the query. The basic logic: if every page on page one of the SERP has hundreds of high-authority backlinks, ranking there requires comparable link strength.

The exact formulas differ by tool:

• Ahrefs KD. Based on the number of referring domains to the top 10 ranking pages, on a 0 to 100 scale.¹
• Semrush KD. Combines link signals (referring domains, follow/nofollow ratio), the median Authority Score of ranking domains, SERP feature presence, and other factors.²
• Moz Difficulty. Moz’s own score in Keyword Explorer, on a 1 to 100 scale rather than 0 to 100. Moz does describe the inputs: Difficulty “takes into account the Page Authority (PA) and Domain Authority (DA) scores of the results ranking on the first page of Google for the given query, as well as modifying intelligently for projected click-through rate”, and the scores “roughly correspond to a weighted average of the PA of the top 10”, with DA, homepages and query term use modifying that average.³

Because the inputs differ, and none of the three publishes the algorithm itself, the same query often shows different difficulty scores across tools. Don’t compare scores between tools; pick one and use it consistently within a workflow.

What does KD not measure?

KD scores typically do not factor in:

• Content quality of the ranking pages
• Search intent of the query
• Topical relevance of the ranking sites
• Brand recognition of the ranking sites
• SERP feature dominance (in most tools)
• The actual algorithmic difficulty in Google’s eyes
• On-page factors, which Ahrefs states outright: “KD does not take into consideration any ‘on-page SEO’ factors”¹

A query can have a low KD but be effectively unrankable because the SERP is dominated by entrenched brands, AI Overviews, or featured snippets that take all the clicks. Conversely, a high-KD query may be genuinely accessible if the existing top results are weak in content quality despite their link profiles.

Underneath all of those sits a limit of unit. KD scores a query in isolation, but nobody ranks a query: a page ranks, and a page covers a subject spanning many related queries. Google resolves a query to the meaning and entities behind it, so the page that takes the position is usually the one covering the subject properly rather than the one built for the string. A per-query score cannot see that. It is why a query can look hard on its own and be reachable as part of a cluster you were going to build anyway, and why the score belongs to a filtering step rather than a decision. See semantic search for how the matching works.

How do you use keyword difficulty sensibly?

As a coarse filter. When working through a list of hundreds of keywords, KD is a fast first-pass filter. Drop everything above a threshold (say, KD 60 for a new site, KD 80 for an established one); investigate everything below. The tools publish their own bands if you would rather start from those: Semrush splits its score into six levels from “very easy” at 0 to 14 upwards,² and Moz says it “generally see[s] keywords getting significantly tougher to rank for at a Difficulty score of around 40”.³

Combined with intent and SERP analysis. KD by itself doesn’t decide; combined with a SERP read (what’s ranking, why, what kind of content, whether AI Overviews are present), it shapes a realistic prioritisation.

Relative to your domain authority. A site with DR 30 should not target many KD 70 queries; a site with DR 70 can. The rough heuristic: stay within 20-30 points of your DR for primary targets, with occasional reaches.

Or let the tool personalise it. Semrush now offers Personal Keyword Difficulty (PKD), which scores how hard a keyword would be for a named domain rather than for an average site, assessing “the thematic relevance between the domain you’ve entered and the keyword”, the competition level within the topic, and metrics for your domain and the competitor domains on the SERP.⁴ It automates what the DR heuristic above does by hand, and it accounts for the topical-authority effect that a global KD score cannot see. The usual caveat applies: the calculation is proprietary and model-derived, so read the number as a better-informed estimate, not a measurement.

As a directional indicator over time. If queries you previously couldn’t rank for now show as accessible at your KD, your domain has matured.

What should you do instead of relying solely on KD?

SERP analysis. Look at the top 10 results for the query. What sites? What content depth? What page types? What’s the lowest-authority result in the top 10? That last point is often the most useful: if a single low-authority page is ranking, the query is more accessible than KD suggests.

Estimate click-through rate realistically. A query with 10,000 monthly searches but an AI Overview, three ads, and a sitelinks pack at the top may produce far fewer clicks than the volume suggests. Effective volume matters more than nominal volume.

Consider topical authority. If you’ve already built strong content around a topic, related queries become easier than their standalone KD scores suggest. The cluster-and-pillar model works partly because authority on a topic compounds.

Look at competitor weaknesses. A high KD query where the top results are outdated, generic, or low-effort is a real opportunity even if the score looks intimidating.

When do KD scores mislead?

Branded queries. “Apple iPhone” has high KD reflecting the dominant link authority of the major retailers and brand pages that hold page one. The SERP is effectively closed to smaller sites, not because of branded intent, but because entrenched competitors have set the floor extremely high.

Queries with few competing pages. New or niche queries may have artificially low KD because there’s not enough established competition for the score to calibrate against. Often these are real opportunities; sometimes they’re queries no one searches for.

Queries dominated by SERP features. A query where AI Overviews answer the question, plus a featured snippet, plus People Also Ask, leaves very little room for organic clicks regardless of how rankable position 1 might be.

YMYL queries. Health, finance, and legal queries are often understated by KD because the difficulty isn’t just about links; it’s about E-E-A-T signals KD doesn’t measure.

Does keyword difficulty tell you anything about AI search?

Not directly. KD estimates the difficulty of reaching the organic top 10. Earning a citation in an AI Overview, or in an assistant’s synthesised answer, is a different problem with different inputs: whether a passage is extractable on its own, whether the claims in it are specific, and whether the source is recognised on the topic. A query can be cheap to rank for and hard to be cited on, and the reverse.

Forward-looking scores have started to appear, though not as like-for-like replacements. Semrush’s Prompt Research report carries Topic Difficulty, which it defines as “a score (0-100%) estimating how competitive it is to appear for a topic, based on the brands most frequently mentioned in AI answers”.⁵ The unit is the thing to notice. KD scores a query and asks whether a page of yours can reach the top ten; Topic Difficulty scores a topic and asks whether your brand gets named at all. One is a question about a page’s position, the other about an entity’s standing, and a figure in either says nothing about the other. It is also early, and it is one supplier’s score over its own sample of AI answers.

So the SERP read still carries the weight: if an AI Overview answers the query completely, a reachable top-10 position may be worth considerably less than the KD score implies, and no single metric will tell you that. For what actually happened after publishing, rather than what was predicted before, see AI visibility measurement.

Building a keyword priority workflow

A practical workflow that uses KD without over-relying on it:

1. Generate a keyword universe. Seed terms, related queries, competitor coverage gaps, autocomplete data. See building a keyword universe for how to do this systematically.
2. Filter by KD. Drop queries above your achievable threshold. Adjust threshold based on your domain authority, and keep it generous: step 4 makes some of what you nearly dropped reachable.
3. Filter by intent. Drop queries whose intent doesn’t match content you can produce or commercial outcomes you care about.
4. Cluster by topic. Group remaining queries into pillar-and-cluster structures. Plan content per cluster, not per keyword.
5. SERP-check the top 10-20 priorities. For each, validate that the SERP is reachable; check for AI Overviews, featured snippets, and SERP features that may suppress clicks.
6. Rank by opportunity score. Some combination of effective volume × commercial value × confidence in ranking. Custom to your business.

KD is one input to step 2. The rest of the workflow is what produces good keyword strategy.

Frequently asked questions

Can low-KD keywords drive meaningful traffic?

Yes, in aggregate. Long-tail strategy depends on this; individual low-KD queries may have low volume but a portfolio of them produces sustained traffic. The key is volume across the portfolio, not per-keyword.

Should I avoid all high-KD keywords?

Not categorically. High-KD queries are often the highest-commercial-value queries. The right strategy is to pursue them through long-horizon content development and link building, not to target them on day one.

Why do tools disagree so much on KD?

Different methodologies, different backlink indexes, different normalisation. Pick a tool, use it consistently, and treat the absolute numbers as relative to that tool’s calibration.

Footnotes

1. What does KD stand for in Keywords Explorer? — Ahrefs Help Center ↩ ↩²

2. Keyword Difficulty score — Semrush ↩ ↩²

3. Keyword Difficulty — Moz Help Hub ↩ ↩²

4. How is Personal Keyword Difficulty (PKD) calculated? — Semrush ↩

5. AI Visibility Metrics — Semrush ↩