SEO tools

These are content analysis tools: they tell you what a page is actually about and how hard it is to read. Two things make them different from the usual free SEO utilities. They do not upload your draft copy, and they do not break when the copy is in Chinese.

Why the standard formulas fail on Chinese content

Almost every SEO content metric was built on English and takes English structure for granted. Keyword density divides by a word count, and word counts are produced by splitting text on spaces. Chinese does not put spaces between words, so the split returns a single token for an entire article: the denominator becomes one, that one "word" gets a density of 100 per cent, and every term you actually wanted to check gets zero — or the tool divides by zero and prints NaN.

Readability is worse, because the failure is silent. Flesch Reading Ease and Flesch-Kincaid divide by syllables per word. Chinese has no syllable count in the sense the formula requires and no whitespace-delimited words to divide by, so the result is a negative number, a NaN, or an absurdly high score for a genuinely difficult text. The tool still shows a number, and a number on a screen gets acted on.

Neither of these is an exotic edge case. It is the default behaviour for every page written in Chinese, Japanese or Korean, and because the output looks like a normal table of percentages, nobody notices that the percentages mean nothing.

What we do instead

The keyword density checker reports two figures with two different denominators. By character divides the characters a term occupies by the total content characters, which behaves identically for both scripts and answers "how much of the visible page is this term". By phrase divides occurrences by the number of phrase positions — word slots for Latin script, n-gram windows for CJK — which is the closer analogue of the conventional percentage. Neither denominator can collapse to one, so neither can produce 100 per cent for a Chinese paragraph.

The readability score reports the three standard formulas for Latin-script text and marks them not applicable when the text is mostly Chinese, rather than printing a fabricated number. For Chinese it reports a separate index built from three things Chinese does have: average sentence length in characters, coverage by common characters, and punctuation density. That index is explicitly labelled as our own, explicitly not comparable with a Flesch score, and shown alongside the three raw figures so you can see what moved it.

Treat both numbers as diagnostics, not targets

Keyword density is not a ranking factor with an optimal value, and the "ideal 2 to 3 per cent" from older SEO writing was never more than a rule of thumb. Its real use is as a mirror: a filler phrase sitting at the top of the list, or your actual subject missing from the top ten, are editorial findings you can act on. A figure above roughly six per cent is worth attention not because an algorithm punishes it but because a human reader notices the repetition.

Readability formulas are equally crude by design. Nothing in them understands your argument or your structure, so a good score means your sentences are short and your words are simple, not that your writing is clear. Both metrics are most useful compared against an earlier version of the same page, and both are actively harmful as optimisation targets — writing to hit a readability score produces text that scores well and says nothing.

Your unpublished copy stays on your machine

Page copy is normally checked before launch, which is exactly when it is confidential: campaign landing pages, client work under an NDA, pricing announcements, competitor material you are analysing. Most free SEO checkers are a form post to somebody else's server, and the copy sits in their logs.

These tools compute in your browser tab. There is no upload request, the site's Content-Security-Policy restricts connections to our own origin, and the pages keep working with the network disconnected. That last one is the easiest proof available: disconnect, reload, paste, and watch the numbers still appear.

What is coming to this category

Meta tag and Open Graph generators, a robots.txt generator and tester, a sitemap generator that works from a list of URLs you paste rather than by crawling your site, an hreflang generator and a schema markup generator are all planned. The crawling ones are deliberately being built to not crawl: a tool that fetches your site from our servers would need the connection badge and would be slower and less private than pasting the list you already have.

For the Chinese market specifically, the more valuable additions are the ones no English-first SEO suite bothers with: full-width and half-width normalisation for meta fields, correct CJK-aware truncation warnings for titles and descriptions measured in rendered width rather than character count, and Chinese-aware slug generation. Those are on the roadmap ahead of another rewrite of a tool that already exists in ten places.