Keyword density is the percentage of total words in a text that a specific term or phrase occupies, calculated by dividing the term's count by the total number of recognized word tokens and multiplying by 100. The formula is straightforward: density = (count ÷ total words) × 100, where total words is every Unicode letter-or-number sequence the tokenizer recognizes, not just the words a tool decides to display in its frequency table. To find keyword density accurately, paste the visible text you want to inspect into a browser-based counter, optionally enter one exact phrase whose repetition you want to track, and let the tool report total tokens, unique tokens, the top frequencies, and an optional contiguous phrase count, all using the same disclosed denominator. The percentages that appear are descriptive evidence about your copy, not a target to engineer toward. A transparent report lets you spot accidental repetition, missing terminology, or editorial imbalance, while a hidden denominator hides the very thing it claims to measure.
What the Keyword Density Formula Measures
Keyword density answers one narrow question: out of every recognized word in your text, what share does a specific term occupy? The formula is straightforward: density = (count ÷ total words) × 100. Both numbers come from the same tokenization pass, which means changing how you count words changes the density. A piece of text with 800 word tokens and a keyword that appears 16 times yields 16 ÷ 800 × 100 = 2 percent. The same 16 occurrences inside 500 tokens would yield 3.2 percent, even though nothing changed about the writing itself. The percentage is therefore a ratio between two counts drawn from the same source, not a property of the writing alone.
The denominator is the part most casual tools get wrong. A frequency report that filters short words or strips common terms before computing the denominator silently inflates the percentages it shows. A transparent tool counts every recognized token for the denominator and only filters which rows appear in the display, which is why the Keyword Density Checker keeps a separate Minimum word length control that affects the table layout but never the underlying total.
Why the Denominator Has to Stay Honest
The denominator is the single number that turns a raw count into a percentage, and a dishonest denominator turns the percentage into fiction. Three behaviors keep it honest across inputs:
- Every recognized token counts. Whether the word appears in the displayed frequency table or not, it contributes to the total the same way.
- Case does not change the count. Red, red, and RED all merge into the same row because matching is case-insensitive through Unicode lowercase conversion.
- Internal apostrophes stay inside the token. Forms such as don't or l'été remain single units, so contractions and accented words count as one word rather than splitting into two or three.
Without those guarantees, two reports on the same paragraph can disagree about the same phrase, which turns any editorial decision built on the numbers into guesswork.
| Tokenizer behavior | What it means for the count |
|---|---|
| Unicode letters or numbers grouped together | Counted as one token |
| Internal straight or typographic apostrophe | Stays inside the token (e.g., don't, l'été) |
| Punctuation, dashes, symbols, whitespace | Treat as separators between tokens |
| Different capitalization of the same word | One merged row, one merged count |
| Short token hidden by Minimum word length | Still in the denominator, only absent from the visible table |
How to Find Keyword Density Step by Step
The fastest path from a finished paragraph to a trustworthy density number is the same path every time: paste, configure, inspect.
- Paste the visible text you want to inspect. Drop the final visible copy into the input area. Avoid copying navigation menus, footers, sidebars, and repeated templates, because those words become part of the denominator and distort the percentages. The checker is bounded at 500,000 characters and rejects empty or token-free content rather than returning a misleading zero report.
- Optionally enter one exact keyword or phrase. Type the contiguous phrase whose repetition you want to track. Tokenization follows the same rules for the query as for the text, so matching is exact and case-insensitive, and phrase matches may overlap.
- Choose the minimum word length for displayed single-word rows. A higher value hides short rows from the table. The setting changes the visible report, not the denominator: total words and density still include every recognized token regardless of the filter, so a minimum of two hides one-character rows but does not pretend those tokens were absent from the source.
- Select Check keyword density. The page then reports total tokens, unique normalized tokens, the 50 most frequent eligible words, and the density for the optional phrase, all using the disclosed denominator.
- Review counts and surrounding prose. Treat the percentages as descriptive evidence about your copy, not as a target. Read the sentences that contain the repeated terms to judge whether the repetition is intentional or accidental, and revise for readers rather than for the metric.
How to Read the Frequency Table
The frequency table is the part most readers skim first, so the columns it shows matter. Three rules govern what appears and how to interpret it.
- Rows sort by descending count, then alphabetically for ties. The top of the table is the most common word in the eligible set; the bottom of the visible portion is the 50th. Sorting is deterministic, so running the analysis twice on the same text produces the same order.
- At most 50 terms display. The interface stays readable on a laptop screen. Total and unique counts still cover the complete input, so the rows you cannot see have not disappeared from the analysis.
- Density is count divided by total tokens multiplied by 100. Single-word density uses that word's count. Phrase density uses the phrase's contiguous occurrence count, and is not divided by possible starting positions or multiplied by phrase length.
Consider the test sentence "Red blue red. RED green blue red-blue." The tokenizer finds eight total words. Red appears four times, so its density is 4 ÷ 8 × 100 = 50 percent. Blue appears three times, so its density is 3 ÷ 8 × 100 = 37.5 percent. Green appears once, so its density is 1 ÷ 8 × 100 = 12.5 percent. The exact phrase red blue occurs twice, producing a phrase density of 2 ÷ 8 × 100 = 25 percent under the documented denominator. Every percentage in the table is count ÷ total tokens × 100 applied to whatever row you are looking at.
Three Common Pitfalls When Interpreting Results
A density number is easy to misread because the math looks simple and the writing context is not.
- Filtering inflates the percentage. If a report removes common short words before computing the denominator, the remaining percentages climb. Compare reports using the same tokenizer and denominator before drawing any conclusion about your draft.
- Repeating a phrase to hit a target makes the copy worse. Search engines evaluate usefulness, intent satisfaction, originality, structure, links, reputation, and many other signals. Adding a phrase to reach a numeric percentage can read as keyword stuffing, and Google's spam policies explicitly call out keyword stuffing as a manipulative practice.
- Comparing competitors' densities without their tokenizers. Two sites can report different percentages for the same paragraph because their tokenizers treat apostrophes, dashes, or accented characters differently. Treat competitor numbers as approximations rather than as targets.
When Density Helps and When It Misleads
Density is most useful when it surfaces something you missed while writing. A term you did not realize you had written eight times, a brand name that appears in two paragraphs but not a third, a section that leans on a single phrase because the surrounding vocabulary is thin — these are the patterns a transparent report reveals. The percentages work as evidence about your draft, not as a verdict.
Density is least useful when it becomes a target you engineer toward, because every word added to push the percentage higher is a word chosen for the metric rather than for the reader. The cleanest workflow is to paste the final visible copy, enter the exact phrase you care about, run the analysis, and read the surrounding sentences rather than acting on percentages alone. Compare revisions using the same tokenizer and denominator so the numbers stay comparable across drafts. Keep human editorial judgment and actual search performance data above any isolated density number, and remember that the percentage is descriptive evidence about your copy, not a prescription for ranking improvement.