Keyword Density Is Not a Ranking Factor: What to Measure Instead

Somebody will tell you the target is 1 to 2 percent. Somebody else will say 0.5 to 2.5. A plugin will turn its light green when you hit whatever number it was configured with in 2012, and a writer will spend twenty minutes wedging a phrase into a paragraph that did not want it.

None of these numbers come from Google. There is no keyword density target, there has not been one since the mid-2000s, and John Mueller has said so in public more times than seems necessary. The metric survives because plugins keep displaying it, and a number on a screen feels like measurement.

The number is still worth looking at. Just not for the reason anyone thinks.

---

Where the Idea Came From

Early search engines counted terms. A page mentioning "toronto plumber" nine times looked more about Toronto plumbers than a page mentioning it twice, and for a few years that was genuinely most of the algorithm.

Keyword stuffing killed it. By the time Google had spam detection worth the name, term frequency in isolation was worthless as a ranking signal, and the systems that replaced it worked on relationships between terms rather than counts of them.

What remains today is a family of retrieval and language models that assess whether a document answers a query. Repetition contributes nothing to that assessment beyond confirming the topic, which the first occurrence already did.

What Density Still Tells You

Two things, both diagnostic rather than prescriptive.

It tells you when you have written about the wrong thing. This is the genuinely useful case. You set out to write about commercial roof repair, and the phrase table comes back with "roof replacement" appearing eleven times and "repair" four. You have written a replacement page. Google will treat it as one, rank it for replacement queries, and the repair traffic you wanted will go elsewhere. Nothing in the draft reads wrong; the emphasis simply drifted while you were writing.

It tells you when you have overdone it. Above roughly 3 or 4 percent, text stops sounding like a person. Treat that as a readability threshold rather than a penalty threshold. It costs you through bounce rate and conversion, not through ranking. If a phrase appears so often that a reader notices the pattern, cut it.

Our Content and Keyword Analyzer reports density for one-word through four-word phrases, which is where the drift shows up. Single-word counts are noise. The three and four-word tables are the ones that tell you what your page is actually about, and they frequently disagree with the title.

What to Measure Instead

Five things, in the order we check them.

Query coverage. Does the page answer the questions someone asking your target query also needs answered? For "how much does local SEO cost", that means monthly ranges, what changes the price, what is included, contract length, and when it is not worth buying. A page that answers one of those competes with pages that answer all of them.

Entity coverage. Search engines resolve topics through related entities, not repeated strings. A credible page about Google Business Profile optimisation will naturally reference categories, service areas, review velocity, posts, the local pack, and Maps. Not because a tool told you to include them, but because you cannot explain the topic without them. If your draft covers the topic properly, coverage takes care of itself. If it does not, adding the terms will not save it.

Placement of the primary phrase. Once in the title tag, once in the H1, once within the first hundred words, and in at least one H2 where it fits naturally. That is the whole of on-page keyword placement. Everything past it is superstition.

Information gain. The one that actually moves rankings now and the one nobody measures. What does your page contain that the ten pages currently ranking do not? A price you are willing to publish, a process nobody else documents, a result with real numbers, a clear statement of when your service is the wrong choice. If the answer is nothing, no amount of optimisation fixes it, and the honest recommendation is to write a different page.

Readability, for the reader. Discussed below, because the scores are more misleading than helpful.

Publishing consistently and still not ranking for the terms that matter? Get a Free Visibility Audit →

What a Readability Score Really Measures

Flesch Reading Ease, Flesch-Kincaid Grade Level and the rest are arithmetic on two inputs: average sentence length and average syllables per word. That is it. No score in general use assesses whether the writing is clear, logically ordered, or correct.

Which means a page of confident nonsense in short sentences scores better than a precise technical explanation. You can raise any score by breaking sentences in half and swapping "utilise" for "use". The second change is a real improvement. The first is often a real loss, since a qualification that belonged in one sentence now reads as two unrelated claims.

Used as a smell test, the scores are fine. A grade level of 16 on a page for homeowners means you have written for the wrong reader. A page averaging 34 words per sentence is hard going regardless of who reads it.

Used as a target, they degrade writing. We have seen agencies chop every sentence to fifteen words to hit a green light, producing copy with the cadence of a children's book and none of the substance a buyer spending $8,000 a month needs before they call.

The analyzer reports these scores because they are a useful smell test and because clients ask for them. Treat the number as a prompt to reread a section, not as a thing to optimise.

Running the Check Properly

The workflow that produces a useful result takes about ten minutes per page.

1. Analyse the live URL, not the draft. Templates, navigation, sidebars and footers all contribute text. A 600-word article inside a heavy template can have its phrase distribution dominated by boilerplate, and it is the assembled page a crawler receives.
2. Read the four-word phrase table first. This is where you find out what the page is about. If the top phrases are not the topic, stop and fix the draft before looking at anything else.
3. Compare against the top three ranking pages for your target query. Not to match their density, which is meaningless, but to see which subtopics they cover that you skipped. That gap list is the actual optimisation work.
4. Check the title and meta description at real pixel widths. Character counts are wrong because glyphs differ in width. Our SERP Simulator measures both in pixels, shows desktop and mobile truncation, and previews how the result reads when an AI assistant cites it. A title truncated at the wrong word is a click-through problem that no amount of on-page work compensates for.
5. Fix what the gap list says, then leave the density number alone.

The Version of This That Actually Works

A GTA commercial cleaning company came to us with 40 service pages, each written to a 1.5 percent density target by a previous agency. Every page had the phrase in the title, the H1, three H2s and roughly a dozen times in the body. None ranked past position 30.

We rewrote eight of them. The density on the rewrites is inconsistent, somewhere between 0.4 and 1.1 percent, because nobody was counting. What the pages gained was pricing ranges, minimum contract sizes, what is excluded from a standard scope, and a straight answer about when a smaller local operator is the better choice.

Six of the eight reached the first page within four months. The other 32 pages, untouched, sat where they were. The variable was not the keyword count in either direction. It was whether the page contained information a buyer could act on.

What to Take From This

Look at density once, early, to confirm you wrote about what you meant to write about. Then close the tool.

The pages that win are the ones with something in them. No ratio calculated from the text predicts that. The closest thing to a predictive measure is whether a reader who lands on your page can stop looking.

If you want the full on-page sequence, our SEO content guide covers structure and intent matching, and the on-page SEO checklist covers the technical elements around the copy. Google's helpful content guidance is worth reading directly rather than through summaries of it, because the self-assessment questions in it are more concrete than the commentary suggests.

Sitting on dozens of pages that were optimised to a formula and still do not rank?

A Free Visibility Audit identifies which of your existing pages are worth rewriting, which should be merged, and which are competing with each other for the same query.

Get a Free Growth Audit →

---

Related reading: How to Write SEO Content That Ranks | On-Page SEO Checklist | How to Do Keyword Research