Keyword Density Checker

Using the Keyword Density Checker

1. Paste your draft content into the textarea. Plain text is easiest; HTML is stripped of tags before counting, so you can paste a rendered blog post straight from View Source.
2. Optionally enter a specific keyword to spotlight - its exact frequency and density percentage appear at the top of the results.
3. Review the three tables - unigrams (single words), bigrams (two-word phrases), trigrams (three-word phrases). Each row shows raw count and density as a percentage of total words.
4. Read the signal, not the score - the goal is to spot over-optimization. Density above roughly 3-4% on a single phrase is a symptom of unnatural repetition; below 0.5% for your target topic may signal the page does not actually cover the term.
5. Edit and rerun - the count updates on each submission, so you can trim or reword obvious stuffing and verify the density drops.

What This Tool Is For (and What It Is Not)

Keyword density is a diagnostic, not a target. It is useful as a symptom check: a paragraph that mentions "affordable dental implants Chicago" seven times in 400 words has an ~5.3% density that reads as unnatural to any editor and to modern search engines. The tool tokenises your input using String.prototype.split(/\\s+/), lowercases each token, strips punctuation, and builds frequency counts for unigrams, bigrams, and trigrams with basic stop-word filtering. The density formula is straightforward: (occurrences / total_tokens) * 100.

What it cannot tell you is whether your page will rank. Modern search engines moved past keyword-density ranking in roughly 2013 with Hummingbird and have since added BERT (2019), MUM (2021), and passage ranking - all of which model semantic intent rather than literal phrase counts. Use this tool to catch over-optimization and under-coverage; do not use it to hit a magic number. If a density report tells you "add more occurrences of your keyword to reach 2%," that is bad advice - follow the reverse read only.

When to Actually Run a Density Check

• After publishing a page that is ranking for the wrong query - often the density report reveals the page is more about a secondary phrase than the primary target.
• When editing a piece written by multiple authors, to catch inadvertent repetition where each author thought they needed to "establish" the topic.
• Auditing content flagged as "thin" by an SEO crawler (Screaming Frog, Sitebulb) - density combined with word-count gives a fuller picture than either alone.
• Reviewing AI-generated draft content for its tendency to repeat the prompt subject across paragraphs, which inflates unigram density well past natural levels.
• Comparing your draft to competitor pages that rank for the same query - not to match their density, but to spot topic areas they cover that you do not.

Edge Cases That Distort the Number

• Stop words inflate unigram lists - "the," "and," "of" dominate any English text. The tool filters a standard stop-word list, but the filter is language-specific; pasting Polish or Japanese gets unfiltered noise.
• Inflected forms count separately - "implant," "implants," and "implanted" are three distinct tokens to the tool. A light stemmer could merge them, but stemming distorts other statistics; the tool leaves inflection as-is.
• Brand names and product SKUs - counted as ordinary words, which can inflate density for branded content where your product name legitimately appears in every section.
• Hyphenated compounds - "long-tail keywords" counts as one token in most tokenisers and as three in others. This tool treats hyphens as intra-word characters, so "long-tail" is a single unigram.
• Code blocks, URLs, and citations - all counted. If your page has a pasted JSON schema or a reference list with DOIs, those tokens distort the density. Strip them before checking or mentally subtract.
• Very short content - densities on pieces under 300 words swing wildly with single-word edits, which makes the metric unreliable for snippet-length text.

How Modern Search Engines Actually Read Text

Google integrated BERT into Search in October 2019 and extended it globally that December. BERT is a transformer-based language model that analyses the relationship between words in a query rather than matching them as a bag of words. MUM, announced in May 2021, added cross-lingual and multimodal understanding, and passage ranking lets Google rank individual sections of a long page. Together these mean that repeating a keyword in every paragraph no longer correlates with relevance - what matters is whether the surrounding content actually discusses the concept the keyword represents. Google Search Central documentation on content quality explicitly warns against "writing lots of text on popular topics, hoping that some of it will perform well in search results."

Alternative Approaches to the Same Question

For "is my page well-optimized for query X," topical coverage is a better signal: does your page discuss the entities, subtopics, and related questions that top-ranking pages discuss? Tools like Clearscope, SurferSEO, and MarketMuse build topic models from SERP data and score drafts against them. For "is my page over-optimized," a density check is actually helpful, but so is a human re-read. For free alternatives, word-freq in Node.js or Python\'s collections.Counter give similar output. This tool wins for local analysis without uploading drafts to a third-party SEO service.