Keyword Density Analyzer

I once reviewed a client's product page that had rewritten the target phrase 'buy organic coffee' nine times in a 320-word paragraph. The density came out at 5.6%, which is well above the 1–3% range I consider safe for ecommerce copy. I showed them the number, they cut it to three mentions, and the page recovered from a manual action within two weeks. That is the tool's real strength: it gives you an objective figure to point at when someone has overdone it. It is also useful for teaching beginners. When I train junior writers, I ask them to run their first ten drafts through an analyser so they get a feel for natural frequency. Most start at 4–5% and gradually land around 2% without forcing it. The tool works best on short text – think meta descriptions, product titles, alt text. For a 160-character meta description, a single mention of the keyword is about 2.5% density. Two mentions push it to 5%, which is risky. The analyser catches that instantly. I also use it during content migrations when I need to check whether an old page was stuffed before I decide to rewrite or redirect. On one migration, I found a page with 7.2% density for 'used cars london'. I rewrote it to 1.8% using synonyms like 'pre-owned vehicles' and 'second-hand cars'. The page kept its rankings. That confirmed my rule of thumb: for short commercial pages, keep exact-match density under 2.5%.

The moment you feed it a 3,000-word guide, the density drops to something like 0.4% even if you used the exact phrase a dozen times. That is mathematically correct but practically useless. What matters for long content is term saturation across semantic clusters, not a single repeated string. I made that mistake early on: I tried to keep density above 1.5% in a long pillar page and ended up with awkward repetition that hurt readability. The tool cannot distinguish between a keyword used in a heading, a body paragraph, and a caption. Those positions carry different weight. It also ignores synonyms and related entities. You could write a perfectly comprehensive article about 'digital marketing strategy' with a 0.8% density because you used 'online marketing plan' and 'digital promotion approach' instead. The analyser sees low density and flags it as under-optimised, but search engines see a well-covered topic. Another awkward scenario: voice search optimisation. People query 'What is the best coffee shop near me?' not 'best coffee shop near me coffee shop open now.' A density analyser cannot guide conversational phrasing. For the same reason, it is no help with FAQ content that uses long-tail question forms. I have also found it misleads on single-page apps where text is loaded dynamically. Some analysers only count the HTML visible at load, missing chunks rendered by JavaScript. You get a false low number and waste time adding more keywords to an already stuffed page.

For quick checks, I still use seobility's keyword density checker because it shows the formula and is free with no sign-up. But I would not base any decision on it alone. [Google Trends](/google-trends/) gives a better picture of whether a term is actually searched, and that is more important than density. If no one searches your phrase, hitting 2% density is pointless. For my own sites, I rely on [Semrush](/semrush/) for content audit reports that track keyword usage across a full page including headings and image alt text. It also shows related terms, which a density analyser cannot. When I teach workshops, I recommend [Ahrefs](/ahrefs/) because its content gap feature highlights missing related keywords better than any density number. For ecommerce clients, I sometimes use [eRank](/erank/) because it has a dedicated keyword density tool that handles product descriptions well. But if I want to understand how a page actually performs in search, I skip density tools entirely and look at Google Search Console for query data. That tells me which terms users are finding my page with, which is the real measure of relevance. The density analyser is a crutch for beginners. Once you know what natural language looks like, you do not need it. I still use it for the first pass on short copy because it is fast, but I never optimise to a number.