Understanding keyword density
A 1998 SEO trick that's now a smell test.
Why keyword density used to matter, why it doesn't now, and what the n-gram view is still legitimately useful for.
What it measures.
Keyword density = (occurrences of a word ÷ total words) × 100. "Best wireless headphones" appearing 8 times in a 1000-word article is 0.8 % density. Tools report this per word, per 2-gram (consecutive pair), per 3-gram. Originally tracked because 1990s search engines ranked pages partly by how often the query phrase appeared. Stuff "best wireless headphones" 30 times into a page; rank for it.
Why Google no longer cares about it directly.
Modern search engines use neural language models (BERT, MUM, etc.) that understand semantic intent — what a page is about, not which words it mentions most. Density-based SEO advice ("aim for 2.5 % density on your target keyword") is a relic. Stuffing the phrase 30 times now triggers spam-detection penalties more than it boosts ranking. Google's John Mueller has said publicly that keyword density isn't a ranking factor in any meaningful sense.
What it's still useful for.
As a writing smell test. If "blockchain" appears 25 times in a 1200-word article, the author is either over-mentioning or the article is genuinely about blockchain. Either way, useful to see. The n-gram view (top 2-word and 3-word phrases) gives a fast summary of what the page is about — a useful input to the question "is this page focused?". Use the tool diagnostically, not prescriptively.
A worked analysis.
A 1500-word article on cycling. Top 1-grams: "bike" (32), "rider" (18), "tire" (15) — looks like a cycling article. Top 2-grams: "road bike" (12), "tire pressure" (8), "carbon frame" (7) — narrows to road cycling equipment. Top 3-grams: "tire pressure for road" (4), "carbon frame is" (3) — getting specific. The view tells you what topics dominate; if the article claims to cover both road and mountain biking but the n-grams are 80 % road, you know it's actually a road-biking article that mentions mountain biking briefly.
N-gram summary
1-, 2-, 3-word frequencies
Each level reveals a finer-grained topic profile.
Top 1-grams + 2-grams + 3-grams of a 1500-word article
= Topic focus revealed
What real SEO advice looks like.
Write for the reader, not the algorithm. Cover the topic comprehensively — if your competitors' top-ranking articles answer ten related questions, yours should too. Use the target phrase naturally in the title, H1, opening paragraph, and a meta description. Get external links to the page from relevant sources. Make the page fast and mobile-friendly. Density is not on the list; none of these are about hitting a magic word-count ratio.
The legitimate adjacent tools.
Topic modelling (LDA, BERTopic): infers the topics in a corpus from word co-occurrence. TF-IDF: weights terms by how distinctive they are to a document relative to a corpus. Semantic similarity: compares a page's embedding to the query's embedding. All of these are descendants of "what words does this text contain" but operate at a much higher level than raw density. For content strategy at any scale, those are the right tools; keyword density is the entry-level introduction.