SEO keyword density report

Paste the page text, enter one target word or phrase, and choose the percentage that should trigger a warning. The report returns the exact occurrence count, the share of the text occupied by that keyword, and an over-optimization flag. It works locally and deterministically, making it suitable for editorial checks, publishing workflows, and repeatable quality-control rules without downloading a page or guessing what its visible text contains.

Measure the target phrase consistently

Keyword reviews often become subjective because one editor counts exact phrases, another includes partial words, and a third estimates frequency by scanning the draft. This report uses one stable rule. It normalizes letter case and compatible Unicode forms, splits both the page text and target keyword into words, then looks for the complete target sequence. Searching for “shoe” will not count “shoes,” and searching for “running shoe” only counts those adjacent words in that order. Punctuation between words does not prevent a match because punctuation is not a word. Matches may overlap, which matters for repeated constructions and keeps the count mathematically complete. The returned occurrence count therefore answers a precise question rather than approximating one. The report also returns total_words so reviewers can understand the denominator instead of seeing an isolated percentage. Supply the final visible copy when possible: navigation labels, hidden markup, and content omitted from the input cannot influence the result. Because the capability analyzes supplied text rather than fetching a URL, the same input always produces the same answer and remains easy to reproduce during review.

Understand density and the warning threshold

Density is calculated as the number of words occupied by matching keyword occurrences divided by the total number of words, multiplied by one hundred. A three-word phrase found twice therefore contributes six keyword words to the numerator. This definition makes phrase density comparable to single-word density and avoids understating a long phrase merely because it is reported as one occurrence. The result is rounded to two decimal places for a readable editorial signal. Set threshold to the maximum percentage your workflow permits; it defaults to three. The over_optimized flag becomes true only when density is above that threshold, not when it is exactly equal. That boundary makes automated acceptance rules predictable. A warning is a prompt for human review, not proof of a ranking penalty. Search engines do not publish a universal ideal density, and natural repetition varies with topic, format, brand names, and technical vocabulary. Use the threshold as an internal consistency rule, compare similar page types, and read flagged passages before editing. Removing every repeated phrase mechanically can make accurate writing less clear.

Add the report to an editorial workflow

Run the report after substantive editing, when headings, calls to action, captions, and body copy are close to their published form. A content team can store the count, percentage, threshold, and flag beside each draft, then rerun the same request after revisions to show exactly what changed. Agencies can apply different thresholds to product pages, glossaries, and long-form articles without changing the underlying calculation. Developers can also place the capability in a prepublication check: submit the text and focus keyword, inspect over_optimized, and route a flagged draft back to an editor rather than blocking it without explanation. The raw fields make the decision auditable and allow a user interface to display the evidence. Empty text is valid and reports zero words, zero occurrences, and zero density, while an empty keyword is rejected because there is no meaningful target to measure. The algorithm uses no network requests, randomness, clock, or model, so batch results remain deterministic. Each API request costs $0.002; the browser implementation can use the same pure calculation for immediate feedback before automation is introduced.