Keyword Density: Does It Still Matter for Rankings?

Keyword density — the percentage of a page's words that are your target phrase — is SEO's most durable zombie metric. It had a real birthday (primitive engines genuinely counted term frequency), it has been functionally dead for over a decade, and it still headlines tool dashboards and agency checklists that prescribe "2–3% density" like a vitamin dose. Time for the full autopsy: where the idea came from, why it stopped mattering, what actually replaced it, and the one residual way frequency can still hurt you.

Where the number came from

Early information retrieval really did rank partly on term frequency — TF-IDF-family maths where more occurrences (tempered by how common the word is generally) meant more relevance. In the late-90s web, repeating a phrase genuinely lifted rankings, an exploit so trivially gameable that white-text-on-white-background stuffing became the era's signature spam. Density percentages entered the folklore as the "safe" way to play that game — enough repetition to score, not enough to trip filters. The folklore outlived the game by twenty years.

Why it stopped mattering

Three generations of change, each fatal on its own:

1. Statistical spam detection learned unnatural frequency distributions early — density chasing became self-defeating almost immediately.
2. Semantic matching (entity understanding, synonym normalisation, and eventually neural language models) meant pages rank for phrases they never contain verbatim — the system reads meaning, which is the subject of the semantic SEO guide next in this series. When "how to get more sites linking to me" matches a page titled "link building", counting exact strings is over.
3. Google said so, repeatedly — search spokespeople have dismissed density as a factor in plain words for years. Rare is the SEO question with this unanimous an answer.

What replaced it: coverage, not frequency

The legitimate instinct inside density — "the page should be about the topic, measurably" — matured into coverage: does the page address the subtopics, entities and questions a competent answer includes? A guide to backlink audits mentions Search Console, referring domains, disavow, anchors — not because a tool prescribed co-occurrence percentages, but because a complete answer can't avoid them. That's why the working checklist is placement (a handful of structural confirmations) plus completeness, with frequency left entirely to natural prose. Write to cover; never write to count.

The residual truth: stuffing still hurts

Density has no reward curve left, but it retains a penalty cliff. Keyword stuffing is a named spam policy — unnatural repetition, lists of variants, hidden text — and pages that read like "best budget laptops for students who want the best budget laptop" earn suppression from both algorithms and human bounce rates. The tell isn't a percentage; it's prose no human would write. Read your page aloud: if the phrase clangs, cut occurrences until it doesn't. (The same repetition maths applies off-page too — the anchor distribution rules are density's one living descendant, because anchors are a place where unnatural frequency still gets counted.)

What to do when a tool prescribes a density

Content tools that grade "keyword usage: 4/10, add 6 more mentions" are pattern-matching against pages that rank — correlation dressed as instruction. Use their topic and entity suggestions (that's the coverage signal, genuinely useful) and ignore their occurrence arithmetic. If following a tool's counter ever makes a sentence worse, the sentence was right and the counter was wrong — the merit principle outranks every dashboard.

Frequently asked questions

So there's literally no ideal keyword density?

None — no percentage, no range, no per-500-words quota. Pages rank at 0.2% and at 3% alike, because the number was never in the equation. Placement checklist + complete coverage + readable prose is the entire modern formula.

Can too FEW mentions hurt?

Only in the degenerate case where the page never plainly states its topic — which the placement checklist's title/H1/first-paragraph confirmations already prevent. Past those, semantic matching carries the rest.

Why do high-density pages sometimes rank #1 then?

Because density is ignored, not punished-below-the-cliff — those pages rank on links, coverage and intent-fit despite the repetition, not through it. Copy their link profiles and their completeness, never their tics. The next myth in the queue gets the same treatment: LSI keywords. And the ranking inputs that never went out of style live where they always did — earned authority, ours to build with you.