Keyword Density, Prominence and Proximity: What Survived

Many SEO briefs still ask writers to hit a keyword density percentage, place related terms within a certain number of words, and repeat the target phrase near the top of every section. That process creates pages that sound engineered rather than useful, while giving stakeholders a false sense of control.

Of the three classic ideas—density, prominence and proximity—only prominence still has a defensible role as a practical editorial principle. The better replacement for keyword density targets is a combination of entity coverage, intent match and clear information architecture.

The short answer: what survived?

Why keyword density became a bad target

Keyword density is usually calculated as the number of times a phrase appears divided by the total number of words, multiplied by 100. A 1,000-word article containing “keyword density” 20 times would have a density of 2%.

The arithmetic is straightforward. The conclusion is not. A 2% result does not prove that the page is relevant, comprehensive or well written. It also does not tell an editor whether the phrase appears in a useful context. The same percentage could describe a strong guide, a repetitive page or a block of awkward copy written to satisfy a spreadsheet.

Density targets also break down across different query types. A product page may need the product name frequently because it is the subject of every specification. A technical guide may need the phrase only a few times because the surrounding terminology makes the topic unambiguous. A short service page can look “over-optimised” at the same percentage as a long article simply because the denominator is smaller.

There is no responsible universal threshold such as 1%, 2% or 3%. Treat those numbers as workflow preferences at most, not search performance laws. A page that repeats a phrase 30 times to reach a target is more likely to create a poor reading experience than one that uses it 8 times precisely.

The popular advice that is wrong

“Use the exact-match keyword once every 100 words” is popular advice, and it is wrong as a general rule. It assumes that search systems need repeated strings before they can identify a subject. It also encourages writers to alter perfectly good sentences, use unnatural pluralisation and restate the same claim.

Do not add the target phrase to every heading. Do not place it in image alternative text when the image does not communicate that phrase. Do not hide repetitions in tabs, footers or off-screen text. Do not produce separate pages for minor wording variations when one page can satisfy the same intent clearly.

Prominence still has a useful role

Prominence means placing an important term or concept where readers and systems can quickly understand its role. It is not a demand to repeat a phrase aggressively. It is an information-design principle: the main subject should be apparent early and reinforced in the page’s most meaningful structural locations.

For a page targeting keyword density, sensible prominence might include:

• A title that states the subject in normal language.
• An opening paragraph that confirms what the page covers and who it is for.
• A relevant primary heading and descriptive subheadings.
• Natural use of the topic in page copy, image descriptions where applicable and internal link text.
• Structured details such as product attributes, service areas, authorship or dates when those details genuinely exist.

Prominence is about priority, not frequency. If a page is about measuring keyword density, readers should not have to infer that fact from paragraph 6. Conversely, using the phrase in the title, introduction and one or two later passages may be enough if the page then discusses formulas, limitations, editorial alternatives and practical checks.

A useful editorial threshold is to make the primary topic unmistakable within the first 100 words, then let the rest of the page develop the subject naturally. This is a readability and clarity standard, not a claim about a fixed ranking signal.

Why proximity is not a ranking formula

Proximity describes how close words appear to one another. The idea has a reasonable linguistic basis: “accounting software for restaurants” communicates a different relationship from a page that mentions “accounting” in one paragraph and “restaurants” 500 words later.

The problem begins when proximity becomes a mechanical rule, such as keeping two terms within 5 or 10 words. Language does not work that way consistently. A sentence may need a qualifier between two concepts. A comparison may need several paragraphs to explain why one product fits a particular use case. Forcing terms together can make the writing less precise.

Use proximity to improve meaning, not to manufacture a signal. Keep a product and its defining feature together when that helps the reader. Put a qualification next to the claim it limits. Place a location beside a service when the page genuinely offers that service in that location. These are sound copy-editing choices whether or not a search engine is involved.

There are also cases where separation is useful. A page comparing two pricing models should not compress every related term into one sentence. A legal or medical explanation may need definitions, exceptions and evidence between the initial concept and its conclusion. A fixed proximity target would reward compression when the reader needs structure.

Replace density with entity coverage

Entity coverage asks whether the page includes the important things associated with the topic, not merely whether it repeats the topic’s label. Depending on the query, those things can include brands, people, materials, features, processes, constraints, destinations, standards, costs and alternatives.

For example, a page about choosing project management software might need to cover users, team size, task management, dependencies, reporting, integrations, permissions, implementation and pricing model. Repeating “project management software” 25 times does not compensate for omitting the decision criteria a buyer actually needs.

Build an entity map in three passes:

1. Extract the task. Identify what the searcher wants to do, such as compare options, solve an error, estimate a cost or select a provider.
2. List required concepts. Include the objects, attributes, processes and constraints needed to complete that task. Use customer interviews, sales calls, product documentation and search-result patterns as inputs.
3. Assign each concept a job. Decide whether it belongs in the answer, a comparison table, a definition, a worked example, a warning or a related page.

Coverage does not mean adding every associated word. An entity is valuable when it helps explain the subject or complete the user’s task. A page about a narrow software feature does not need a history of the entire category simply because competing pages mention it.

Intent match is the real quality test

Intent match asks whether the page gives the right answer in the right format at the right depth. A query containing “best” may need a comparison and selection criteria. “How to” generally needs a sequence of steps. “Price” needs cost drivers, ranges or a method for obtaining a quote. “Near me” needs genuine local availability rather than a location keyword inserted into generic copy.

Assess intent using a simple four-part check:

• Task: What decision or action should the reader complete?
• Format: Would a guide, table, calculator, category page or service page serve that task best?
• Depth: What must be explained before the reader can act confidently?
• Evidence: Which claims need examples, specifications, dates, qualifications or limitations?

Score each section from 0 to 2 for usefulness: 0 if it does not address the task, 1 if it mentions the issue, and 2 if it enables a decision or next step. A page averaging below 1.5 needs a content improvement before an editor spends time adjusting phrase frequency. This is an internal prioritisation threshold, not a search-engine metric.

A practical workflow for editors and SEO leads

Start with one primary phrase and a short list of variants that reflect how customers speak. Do not create a separate requirement for every grammatical form. Then write a one-sentence promise describing what the page will help the reader do.

Next, create a coverage brief with three columns: required concepts, evidence or examples, and planned location. A 1,500-word guide might contain 8 to 12 major concepts, with the most important three introduced in the first 300 words. The exact count should follow the topic; it is not a quota.

After drafting, run three separate reviews:

1. Prominence review: Can a reader identify the subject, audience and purpose from the title, introduction and headings within 30 seconds?
2. Coverage review: Are the necessary entities, attributes, alternatives and limitations present?
3. Language review: Does every repetition sound natural, and can any sentence lose the phrase without losing meaning?

A focused review usually takes 30 to 90 minutes for a standard commercial page. For budgeting, a specialist content audit may fall around £300 to £1,500 depending on the number of pages, research depth and technical access required; those are planning ranges, not market-wide statistics. A small site can often fix its highest-value pages in 2 to 4 weeks, while a large site may need a quarterly prioritisation cycle.

What to report instead of a density score

Remove keyword density from the headline performance section unless the purpose is to demonstrate that no target was used. Report measures that connect to editorial decisions: primary-intent coverage, unanswered subtopics, clicks to relevant pages, qualified conversions, engagement with comparison or pricing sections, and the number of pages competing for the same task.

For a content brief, replace “use the keyword 15 times” with instructions such as “explain the three main cost drivers,” “compare the two common implementation approaches,” or “define the limitation before recommending the method.” Those instructions are testable and useful to a writer.

The surviving principle is simple: make the important subject prominent, then earn relevance through complete, accurate and intent-matched coverage. Keyword density can remain as a descriptive calculation for auditing repetition, but it should not control the copy. Proximity can guide clear sentences, but it should not become a word-distance game. The page wins when the reader can understand the topic, trust the explanation and complete the task without being forced through artificial repetition.

Related reading

• AI search optimisation — scope, method and what we measure.
• Pricing
• What Makes a Website Bad for Search, Measurably
• Conversion Rate: Defining the Denominator Before Optimising