SEO Book's keyword density analyzer: it asks for a free account now, it runs over plain HTTP, and the method on its own page is the part worth keeping

SEO Book's keyword density analyzer is still live, at tools.seobook.com, and two things about using it have changed since it was the default free density checker. First, it no longer runs for an anonymous visitor: the page states that account registration is now required, and gives its reason in its own words, "Someone was effectively DDoSing our server by hitting this tool thousands and thousands of times. Thus we are now forced to require account registration to use this tool." Registration is free. Second, the tool answers over plain HTTP only. Read on 20 October 2026, the http address returned the page and 29.6 KB of it, while the https address failed to connect at all. Everything below was read from that page on that date.

What the page itself says about density, which is the useful part

Most pages about keyword density quote a target percentage. SEO Book's own page refuses to, and the refusal is the strongest thing on it: "There is no single optimal or universal keyword density percentage." It gives the reason, which is that search engines compare or normalise a document against the other top documents for that query, so the threshold is a property of the query rather than of your page. It adds that a two word phrase such as credit cards naturally appears as a phrase while other terms spread out, and that a trusted site with strong links and usage data can carry more repetition than a newer one.

It then gives a frequency rule of thumb in two halves, which is a genuinely useful way to think about it:

• In a trusted corpus of internal content, such as a site's own search index or a set of known authors, higher frequency is generally better.
• In a broad external corpus such as general web search, where people have an incentive to game the result, less is generally better.

And it points at the enforcement history: when the first Penguin update rolled out in April 2012, the page says, on-page classifiers rolled out with it that penalised pages with excessive word repetition. It also notes that Google's rater guidance tells raters to score low-information repetitive content poorly.

The comparative method the page prescribes

Instead of a target, the page gives a procedure, and it is the thing to copy whichever tool you end up using:

1. Search for your target keyword in Google.
2. Take five of the top ranked pages from the results.
3. Analyse each one separately.
4. Weight the smaller, lesser known sites that still rank well more heavily than the big brands, because, as the page puts it, highly trusted brands can rank more on brand strength than on the page itself.

That last instruction is the one most people skip and it is the one that makes the exercise honest. If your site is six months old, the national retailer ranking second is not a model you can copy. The small competitor ranking fourth is.

Record it like this, so the output is a range rather than a single number to chase:

Column	What goes in it
URL	Each of the five ranking pages, plus your own on the last row
Site size and age	A rough note, so you can discount the brands as instructed
Exact phrase count	Whole phrase matches only
Total words	The article text, excluding menu and footer
Density	Count divided by total words, times 100
Placement	Is the phrase in the title, the first paragraph, a heading

Compare your own row against the spread of the other five. A page sitting well above all of them has something to fix. A page sitting at zero has something more important to fix.

Working over plain HTTP, sensibly

Because the tool is served over http rather than https, a current browser marks the page as not secure, and anything you submit travels unencrypted. The practical handling is simple and it costs nothing: give it a published URL, or paste text that is already public on your site. Keep unpublished drafts, client work under an agreement and anything commercially sensitive out of it, and run those through a local count instead. The arithmetic is three lines of code and the working is set out in the keyword density formula, the calculation and a worked example, including the one trap that makes two tools disagree about the same page.

The counting rule that makes two tools disagree

There are two accepted ways to count a multi word phrase and they produce different percentages for identical text. Count each full phrase as one occurrence, or count the words the phrase consumes. Ten uses of a three word phrase on a 1,250 word page is 0.8 percent on the first method and 2.4 percent on the second. Both get called keyword density. Pick one method, use it for every page you intend to compare, and write down which one you picked, because the moment you compare a number from one tool against a number from another you are reading a difference in method as a difference in your writing.

What else is on tools.seobook.com

The density analyzer sits inside a larger set, and the navigation on that page separates them into public tools and members-only tools. The public group includes a keyword tool, analytics tools, keyword research, competitive research, pay per click tools, link analysis and rank checkers, plus Firefox extensions and a toolbar. The members-only group listed alongside includes local rank, a website health check, a duplicate content checker, a domain keyword tool, a keyword list comparison, an SEO site planner and hub finder. Named here as the site lists them on 20 October 2026, with no pricing restated.

Questions about the SEO Book density tool

Is the SEO Book keyword density analyzer still free?

Yes. The page requires a registered account and the page's own registration form offers that account at no cost. The requirement is new relative to how most people remember the tool, which is why a bookmark that used to load a working form now loads a sign-in prompt.

Why does it ask me to log in now?

The page answers this itself, and the reason is load rather than monetisation: it says the tool was being hit thousands and thousands of times in a way that amounted to a denial of service against the server, so registration became a condition of use.

Does it work over https?

Read on 20 October 2026, no. The http address returned the page and the https address did not connect. Treat the input box accordingly and give it published text.

What keyword density does SEO Book recommend?

It declines to give a number and says there is no single optimal or universal percentage, because engines normalise against the other top documents for a query. Its recommendation is the comparative method above: measure five ranking pages and read your own against their spread.

Is keyword density still worth measuring at all?

It is worth measuring as a smoke alarm rather than as a target. A page measuring far above every ranking competitor almost certainly reads badly, and the fix helps the reader first. A page where the phrase never appears is the bigger and more common problem, and it is fixed by putting the exact phrase in the title, the opening paragraph and one heading.

Which number should I care about instead?

Placement and coverage. Whether the first lines answer the question somebody typed, whether the related terms a real article would contain are present, and whether a person reading the page aloud would stop. Those three decide the page; the percentage is a by-product of them.

What a density figure cannot see about your page

A density count measures how often a string appears in a document. The decision that now sits alongside ranking is a different one: an assistant asked a question about your topic reads for a sentence it can use, and it either finds one or moves to the next source. A page tuned so that a phrase appears often gives it frequency. A page whose opening paragraph answers the question outright gives it something to quote, and that is a separate property you can check directly by asking. AI Knows Us puts the questions buyers type to ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews and records which pages each answer was built from, so the wording you tuned for a search result can be compared with the wording an assistant picked up.