What Is Keyword Density? Why It’s an Outdated Metric and What to Focus on Instead

Keyword density is the percentage of times a specific word or phrase appears on a page relative to the page’s total word count. The calculation itself is simple: divide the number of times a term appears by the total word count, then multiply by 100. If a 1,000-word page mentions “running shoes” 10 times, its keyword density for that phrase is 1%.

The calculation is easy. The harder, more useful question is whether that number should ever guide a writing decision, and the honest answer is: not directly, not anymore, and arguably not for a long time.

Where keyword density came from

In early search engines, ranking systems leaned heavily on raw term frequency because there wasn’t much else to work with. Algorithms had limited ability to understand context, synonyms, or intent, so the literal repetition of a query term on a page was one of the stronger signals available. That environment created an incentive to repeat target phrases as often as possible, which produced the keyword-stuffed pages that defined a lot of early-2000s SEO and that made web search results noticeably worse to read.

Search engines responded by getting much better at evaluating relevance through methods that don’t depend on raw repetition counts. Two developments matter here:

TF-IDF (term frequency-inverse document frequency) is an information retrieval concept that weighs how often a term appears on a page against how common that term is across a whole corpus of documents. A term that’s rare across the corpus but frequent on a specific page is treated as more distinctive of that page’s topic than a term that’s common everywhere. This is a much more nuanced approach than counting raw occurrences, though it’s still fundamentally about word frequency rather than meaning.

BERT, a natural language processing model Google announced in 2019, represented a bigger shift: the ability to understand words in the full context of the words around them, in both directions, rather than processing a query as an isolated sequence of terms. Google’s own announcement explains that BERT models “consider the full context of a word by looking at the words that come before and after it,” which is precisely the capability that makes counting literal keyword repetitions an increasingly poor proxy for relevance.

There’s no official target density, and Google has said so

There is no published, official ideal keyword density. Google’s John Mueller has said on multiple occasions, across Search Console help threads and public Q&A sessions, that there’s no target density to aim for and that focusing on a specific percentage isn’t a useful way to think about writing content. Treat that as a well-documented general position rather than a single quotable line, since the exact wording has varied across the different times he’s addressed it; the consistent substance is that no fixed percentage makes content rank better, and chasing one is wasted effort.

This is also reflected in how ranking systems actually behave: studies and informal correlation analyses across the SEO industry consistently find weak or inconsistent correlation between keyword density and rankings, which is exactly what you’d expect if density isn’t a direct input to ranking at all.

When repetition becomes a problem

None of this means word repetition is irrelevant in every direction, just that there’s no virtue in maximizing it. At the high end, unnatural repetition (keyword stuffing) is still a real issue, just not because it crosses some precise percentage threshold. It’s a problem because it produces unreadable content and because Google’s spam policies explicitly address it. Google’s documentation on spam policies for Google Search defines keyword stuffing as “the practice of filling a web page with keywords or numbers in an attempt to manipulate rankings in Google Search results,” and treats it as a violation regardless of the specific frequency involved.

In practice, the SEO industry commonly cites a rough range of around 3 to 4 percent as an informal heuristic for when density starts to look unnatural or stuffed. It’s worth being precise about what this number is and isn’t: it is not a Google-published rule, it is not a hard ceiling, and it is not something any ranking algorithm is known to check directly. It’s a community rule of thumb, useful mainly as a rough sanity check during a content audit, not a target to write toward or a line that triggers an automatic penalty the moment it’s crossed.

What actually matters instead

Old approach	What to do instead
Hit a target keyword percentage	Cover the topic comprehensively; let term frequency fall out naturally from genuinely explaining the subject
Repeat the exact phrase throughout	Use natural variations, synonyms, and related terms a person would actually use when discussing the topic
Count keyword instances	Check semantic and topical coverage: does the page address the subtopics a searcher would expect?
Match competitor density	Identify what topics or angles competitors cover that you don't, not what percentage they hit
Treat every page the same	Match term usage to what's natural for the content type and intent (a product spec table reads very differently from a long-form guide, and that's fine)

Placement still matters more than raw frequency. Using the primary topic phrase, or a natural variation of it, in the title tag, the H1, the opening paragraph, a relevant subheading, the URL, and image alt text where appropriate gives both readers and search engines a clear, early signal of what the page is about. That’s a structural and placement question, not a counting question, and it has nothing to do with hitting any particular overall percentage.

A practical way to check for stuffing without a formula

Reading content aloud is a surprisingly effective stuffing check: unnatural repetition is far more obvious to the ear than to the eye when skimming text on a screen. If a sentence or paragraph sounds like it’s straining to fit a phrase in again, it almost certainly is, regardless of what the resulting percentage works out to. A genuinely useful audit process looks at this together with a few other signals:

• Does the content read naturally when read aloud, or does phrasing feel forced around a specific term?
• Does the page address the topic with enough breadth that synonyms and related concepts appear organically, rather than just one exact phrase over and over?
• Are structural elements (titles, headers, intro) using the topic clearly without needing the body copy to repeat it dozens of times?
• Does competitor content in the same space cover meaningfully different subtopics, suggesting a coverage gap rather than a density gap?

Tooling

Most SEO content tools and crawlers, including general-purpose platforms like Ahrefs and on-page analysis features in various content editors, will report a keyword density figure if you ask for it, since it’s a trivial calculation. The existence of the metric in a tool’s report doesn’t make it a meaningful optimization target; treat it the same way you’d treat any other vanity metric a tool happens to surface, useful as a rough sanity check against extreme stuffing, not as a number to write toward.

The bottom line

Keyword density was a reasonable proxy for relevance when search engines had little else to go on. It stopped being one once ranking systems gained the ability to understand semantic meaning, synonyms, and topical context, well before BERT, and even more so after it. There is no ideal percentage to hit, and writing toward one, rather than toward genuinely thorough, naturally worded coverage of a topic, is effort spent solving a problem that doesn’t really exist anymore. If you’re auditing existing content, skip the density calculator entirely and read the page aloud instead, that single check catches what the percentage never could.