Free Keyword Density Checker: Top 100 Phrases on Any URL

Paste a URL and you get a table of the 100 phrases that page repeats most, ranked by how often each one appears. Every row carries four things: the phrase, its raw occurrence count, its density as a percentage, and two flags showing whether that phrase turns up in the page's title tag and meta description.

There is no field for a target keyword, and that absence is the design. A checker that asks what you want to rank for can only ever confirm the belief you brought with you. This one reads the page and tells you what it says. If the phrase you had in mind isn't in the top ten rows, that is the result — not a configuration problem.

What gets counted is the visible text of the page as a crawler would encounter it. We follow redirects up to five hops so you measure the page that finally renders, then strip <script>, <style>, <noscript> and <template> blocks before reading the body. Navigation, footers and boilerplate all count, because they are part of what the page says about itself — a site whose footer repeats its city name forty times has genuinely made that a prominent phrase, whether or not it meant to.

Three lengths are counted in the same pass: single words, two-word phrases and three-word phrases. Single words tell you the topic. The multi-word rows tell you the angle, and they are usually the more interesting half of the table.

Density is a fraction, and the argument is always about the denominator.

Before counting anything, this tool discards words shorter than three characters, bare numbers, and 43 common stopwords — the, and, with, your, from and their relatives. What remains is the page's meaningful vocabulary, and that is what every percentage is divided by. A checker that counts every word on the page divides by a substantially larger number and reports a smaller percentage for the same phrase on the same page.

Here is the gap, measured rather than asserted. On 8 August 2026 we ran Moz's own "What Is SEO?" explainer through this tool. The word seo appeared 97 times. What that is worth as a percentage depends entirely on what you divide it by:

Same page, same day, same word. Two legitimate numbers 1.56 percentage points apart — on a scale whose entire folklore lives between 1% and 3%.

This is why a density figure quoted without the name of the tool that produced it means nothing, and why "aim for 2%" is advice that cannot survive contact with a second tool. Compare pages within one checker, or don't compare at all.

The other thing that run shows is worth sitting with: 5.06% is above the 5% that conventional wisdom calls keyword stuffing, and the page in question ranks at the top of one of the most competitive queries in the industry. The threshold isn't a threshold.

Keyword density was a real ranking input on search engines that no longer exist. Early crawlers had little more than the words on a page to work with, so counting them was a reasonable approximation of what a page was about. That approach was gamed within about a year, and the industry has been running on the residue of it ever since.

Google's position has been consistent and public for well over a decade: there is no keyword density ranking factor, no optimal percentage, and no benefit to repeating a phrase more often. Modern retrieval works on meaning. A page can rank for phrasings it never uses verbatim, and a page that uses a phrase forty times gains nothing from mentions eleven through forty.

What repetition can still do is work against you, and this is the part worth taking seriously. Text written to hit a number reads like text written to hit a number. Content produced primarily for search engines rather than for people is the explicit target of Google's spam policies and its helpful-content signals. So the practical risk of density thinking isn't a penalty for crossing a percentage — it's that optimising toward a percentage produces exactly the kind of writing those systems are built to demote.

One idea worth retiring while we're here: LSI keywords. The advice to "include LSI keywords" appears in almost every article on this topic, including the one this page replaced. Latent Semantic Indexing is a real technique from the late 1980s, and Google has publicly stated it does not use it. There is no list of LSI keywords for your topic, because there is no index they would be keys to. Writing naturally about a subject will produce related vocabulary on its own; that is a description of good writing, not a technique.

Start with the count column, not the density column. The count is a fact about the page. The percentage is a fact about the page and about our tokeniser, which makes it useful for comparing rows to each other and useless as an absolute target.

Then read the top three rows against the In title column. This is the ten-second version of the whole exercise. A page that repeats something thirty or forty times and never says it in the title tag has a mismatch, and the title is almost always the thing that's wrong. It's a one-line change with more effect than any amount of body editing.

Then check In description. The phrases a page genuinely repeats are the raw material for a meta description that matches it. If none of the top ten appear there, the description was written about a different page — or written before the page was.

Then look at the shape of the drop. A page about one thing falls away steeply after its first few rows. A page about nothing in particular has a long flat tail where everything sits at similar frequency. Focus is visible in the profile of the numbers, and it's the property that most often separates a page from the one outranking it.

Then read the multi-word rows. In that same Moz run, 24 of the 100 rows were two- or three-word phrases. Those rows are where the specific claim of a page lives. Single words rarely distinguish two pages competing for the same query; phrases do.

A percentage calculated over a very small number of words is arithmetic, not information. On a 200-word page, a single extra mention moves the density by half a point. Two extra mentions and you've crossed the entire range people argue about.

This matters more than it sounds, because thin pages are common. Across the sites we audited in the first half of 2026, the Content category failed on 41.7% of them, and the single most common failure inside it was thin pages under 300 words left indexable.

So check the analysed word count above the table before you read anything below it. If it's in the low hundreds, the page has a length problem, and density is the wrong question to be asking about it. Fix the page first; measure it after.

The mirror image is worth noting too. Very long pages dilute everything — a 6,000-word guide will show its central subject at well under 1% simply because there are so many other words. That is not a page failing to signal its topic. It's a page with a lot of words.

This is the finding the tool exists for, and it has three common causes.

The page drifted. A brief said "pricing" and the finished article spends most of its length on integrations, because integrations were the interesting part to write. The fix is structural: cut, reorder, or accept that the page is now about integrations and retitle it accordingly.

The template is louder than the content. Site-wide navigation, a large footer, a repeated CTA block or a sidebar can out-shout a short article. If your top rows are your own product name and menu labels rather than your subject, you're measuring the template. Run a longer page on the same template to confirm, and take the finding to whoever owns the layout.

The page is about the right thing in the wrong words. Your subject is present but split across five near-synonyms, none of which reaches the top ten individually. This is usually fine — meaning-based retrieval handles synonyms without help — but it's worth knowing when a title tag has to pick one of them.

Notice that none of these three fixes is "add more mentions". Adding repetitions to raise a number is the one change that reliably makes a page worse: it degrades the writing, it doesn't affect ranking, and it moves the page toward the pattern spam systems are looking for.

Keyword stuffing is not a percentage. It's an intent — text arranged to be counted rather than read — and it's recognisable on sight.

The patterns that get acted on are specific: blocks of city or product names with no sentence around them, the same phrase forced into consecutive sentences until it stops parsing, text hidden behind an image or set to the background colour, and keyword-crammed alt attributes and title tags. What these share is that they exist for a parser and not for a person. None of them is defined by a density threshold, which is precisely why a density checker can't reliably detect stuffing and why a human reader can spot it in a paragraph.

The practical test needs no tool at all: read the passage aloud. If you wouldn't say it that way, don't publish it that way.

Our guide on how to avoid keyword stuffing goes through the specific patterns in more detail. What the density table adds is a rough early warning — a single phrase towering far above everything else on a page is worth a look, not because the number is too high, but because that shape often accompanies writing that has gone stiff.

Density is a diagnostic you run after writing, in the same pass as your other on-page checks, and it takes about a minute.

1. Run the URL and read the analysed word count. Under a few hundred words, stop and fix the length.
2. Read the top ten rows. Is the page's subject the top row, or at least near it?
3. Check In title and In description on those rows. Fix the tags before you touch the body.
4. Note the multi-word phrases — they tell you the angle the page has actually taken, which is what you'd compare against a competitor.
5. Run the page currently outranking you and compare the shape of the two tables, not the two percentages.

From there the next checks are structural rather than lexical: confirm the title and meta tags are what you think they are, check where the page actually ranks, and run a full site audit if the thin-page problem turns out to be site-wide rather than local to one URL.

And then stop. Density is one cheap signal among many, it answers a narrow question well, and it has no more to give once you've asked it.