Keyword density is the number of times a target word appears on a page, divided by the total word count, shown as a percentage. It got popular around 2008 because it was easy to calculate. It's a poor guide today. Search engines rank on meaning and usefulness, not on how often a word repeats.

Two pages can have the same density and rank very differently. One answers the question in plain language. The other repeats the phrase and says nothing. The metric can't tell them apart. So the real question isn't "what density should I hit?" It's "what does this page help a reader do?"

This piece treats keyword density as a failure mode. It shows how pages get ignored, misread or misattributed, and what to check instead. Start with the writing, because the metric is downstream of it.

1. The 2008 method was a convention, not a rule

Around 2008, a common SEO workflow was to pick a keyword, write a page, and count the repeats. The target usually sat between 1% and 3%. That number spread through forums and tool dashboards. No public Google, Bing or OpenAI document gives a recommended density range. It was a habit, not a rule.

Search engines changed how they read pages. Google's "How Search Works" documentation describes crawling, indexing and ranking as separate stages, and says ranking uses many signals. Google's "Helpful content" guidance stresses usefulness to people over word counting. Check the current in-force version of each document.

Easy to compute isn't the same as useful. A page can hit 2% density and still fail to answer the query. A page can sit at 0.4% and answer it well.

2. Density fails a page in three predictable ways

Check each of these before you touch the copy.

Failure mode A: the page gets ignored. The page repeats the keyword but never defines it, never gives an example, and never answers a follow-up question. A reader bounces. Google's "Helpful content" guidance points to this pattern: content made for search engines first tends to underperform. The page isn't penalised. It just isn't chosen.

Failure mode B: the page gets misread. The keyword is ambiguous. "Java" can mean the island or the programming language. "Apple" can mean the fruit or the company. Without context words and entities, the engine may file the page under the wrong topic. Density doesn't fix this. Explicit context does.

Failure mode C: the page gets misattributed. The page copies a phrase so closely that it looks like a duplicate of another source. Google's documented duplicate content handling and its canonicalisation guidance explain how near-duplicates get consolidated. The page loses its own identity, and the original gets the credit.

3. Keyword stuffing is a spam policy issue, not a density number

Keyword stuffing means repeating a term far beyond what a reader needs, often in lists, hidden text or alt attributes. Google's spam policies page names "keyword stuffing" as a spam technique, alongside hidden text and link spam. That page is the authority here. Check the current in-force version.

Google's spam policies describe the likely outcome as ranking lower or being ignored, not a manual action in every case. The policy language is about manipulation, not about a specific density threshold. No published percentage triggers a penalty. Anyone who quotes one is guessing.

Small-scale stuffing looks like this: a paragraph that repeats "strong CRM software" four times in three sentences. A product page that lists 40 city names in the footer. An image alt tag that reads "strong CRM software strong CRM software cheap". Each one signals low quality to a human reader, and Google's guidance treats human usefulness as the standard.

4. The old model and the current one differ on every row

The table below contrasts the two approaches. It's a practical comparison, not a formal standard. Each row names what changed and what to check instead.

Dimension2008 approachCurrent approachWhat to check
Primary metricKeyword density %Query intent matchDoes the page answer the query in the first screen?
Keyword formExact-match phraseRelated terms and entitiesAre synonyms, entities and context words present?
Page structureKeyword in H1, repeated in bodyClear headings, scannable sectionsCan a reader skim and find the answer?
LinksAnchor text with exact keywordDescriptive anchor text, natural linkingDoes the anchor describe the destination?
Success signalRanking position for one phraseImpressions, clicks, citations, conversionsAre people finding and using the page?
Penalty riskAssumed density thresholdSpam policy for manipulationWould a reader call this repetitive?

The table makes the shift visible. In 2008 the metric was a proxy for relevance. Today the proxy is gone, and relevance has to be built into the content itself.

5. Keywords still earn their place in five spots

Keywords help the engine and the reader understand what the page is about. Aim for clarity, not count.

The title element. Google's documentation on title links explains that the title element generates the clickable headline in results. A clear, specific title helps. It doesn't need the exact phrase repeated.

The H1 and section headings. Headings describe structure. A reader scanning the page uses them to decide whether to stay. Use plain words that match the question.

The first paragraph. Put the direct answer early. This helps readers and AI answer engines that pull short passages.

Body text and examples. Use the term where it's the natural word. Then use pronouns, synonyms and specific examples. A page about "keyword density" should also mention "word count", "repetition" and "stuffing", because those are the related ideas a reader expects.

Image alt text. Describe the image. If the image shows a density calculator, say so. Don't stuff the alt with the target phrase.

6. Entities and topic coverage replaced density

An entity is a thing the search engine can identify: a person, place, product, concept or organisation. Google's "Introducing the Knowledge Graph" announcement (2012) describes how the engine maps things and their relationships. Schema.org provides a public vocabulary for marking entities on a page.

Relevance now comes from covering the topic, not from repeating one phrase. A page about keyword density that also covers keyword stuffing, semantic relevance, query intent and AI search is more likely to be seen as a complete answer. That's a coverage judgement, not a density calculation.

AI answer engines add a second reader. OpenAI's documentation for its crawlers and Anthropic's documentation for Claude's web tools describe how those systems fetch and process pages. They pull passages. They don't count keyword frequency. A clear, self-contained paragraph is easier to quote than a keyword-dense one.

7. Replace the density target with eight reader questions

Run this checklist before publishing. It's a set of questions, not a formula.

  • Does the first paragraph answer the main question in plain words?
  • Does the page define each technical term on first use?
  • Are related terms and entities present because a reader would expect them?
  • Can a reader skim the headings and find the answer?
  • Would a reader call any sentence repetitive? If yes, cut it.
  • Does the title describe the page without stuffing?
  • Are image alt texts descriptive rather than keyword lists?
  • Does the page link to a primary source for each factual claim?

The list isn't exhaustive. It's a practical filter. If a page passes it, density is irrelevant. If a page fails it, no density target will save it.

8. Match the fix to the symptom

The right action depends on what's going wrong. Use this short decision guide.

  • If the page ranks but readers bounce, rewrite the opening to answer the question first.
  • If the page is ignored for its target query, check whether the page actually covers the topic, not whether the phrase appears enough.
  • If the page is misread as a different topic, add context words and entity markup.
  • If the page is misattributed to another source, rewrite the copied passages and add a canonical link where appropriate.
  • If the page has been manually flagged for spam, remove the repetitive blocks and review the site against Google's spam policies.
  • If an AI answer engine cites a competitor instead, make each section self-contained and source-backed.

Each symptom maps to a specific fix. None of them is fixed by adjusting density.

Key Facts at a Glance

  • Keyword density is a count, not a ranking signal with a published threshold.
  • Google's spam policies treat keyword stuffing as manipulation, not as a density number.
  • Relevance now comes from topic coverage, entities and query intent.
  • Keywords still help in titles, headings, the first paragraph and body examples.
  • AI answer engines pull passages; clear, self-contained sections are easier to cite.
  • A checklist of reader questions replaces a density target.
Does keyword density still matter for SEO?

As a diagnostic, a little. As a ranking target, no. Google's public documentation describes ranking as a multi-signal process and doesn't publish a density threshold. Treat density as a rough signal of repetition, not a goal.

What is a good keyword density?

There's no published good number. The figure often quoted, 1% to 3%, is a convention, not a rule from Google, Bing or OpenAI. A better test is whether a reader would call the page repetitive.

What is keyword stuffing?

Keyword stuffing is repeating a term beyond what a reader needs, often in lists, hidden text or alt attributes. Google's spam policies name it as a spam technique. Check the current in-force version of that page.

Can keyword stuffing get my site penalised?

Google's spam policies describe ranking lower or being ignored as likely outcomes. A manual action is possible but not automatic. No published density percentage triggers a penalty.

How do I check keyword density?

Count how many times the term appears, divide by the total word count, and multiply by 100. Tools do this automatically. Use the result as a repetition check, not a target.