Why one emoji is 8 characters in most counters
A JavaScript string is a sequence of UTF-16 code units, not characters, and .length counts those units. The family emoji 👨👩👧 answers 8: five code points joined by two zero-width joiners, three of them outside the Basic Multilingual Plane and costing two units each. A reader sees one character.
The unit matching what a reader sees is the grapheme cluster (Unicode Annex #29), counted here with Intl.Segmenter. So the family emoji counts 1, an e with a separate combining acute counts 1 rather than 2, and a CRLF line ending counts 1 per rule GB3. The Code units row shows what a counter calling .length reports instead.
What counts as a word
Any segment the Unicode word-break algorithm marks as word-like, which is not "text between two spaces". "don't" is one word, because an apostrophe between letters is not a break. "state-of-the-art" is four, because hyphens are. An emoji is none. Chinese, Japanese and Thai segment by dictionary, so a sentence with no spaces still counts, where splitting on whitespace reports one word for the whole thing.
Tools disagree by a few percent and none is wrong. Word counts "state-of-the-art" as one; a URL is one word in some tools and five in others. If a count is contractual, use the tool the other party will.
Where the reading and speaking rates come from
238 words per minute is the silent reading rate for English non-fiction from Marc Brysbaert's 2019 meta-analysis of 190 studies; the same review puts fiction at 260 and reading aloud at 183. Speaking uses 150, per the National Center for Voice and Speech. The round 200 or 250 most reading-time widgets use has no source behind it.
SMS segments
SMS uses the 7-bit GSM 03.38 alphabet: 160 characters, covering ASCII plus a set of accented letters, so "café" costs nothing extra and "fête" does not fit. The euro sign, square brackets, curly braces and the backslash cost two units each. One character outside the alphabet, and a curly apostrophe or a single emoji is enough, switches the whole message to UCS-2 and drops the limit to 70. Split messages carry a 6-byte header per part, so they allow 153 rather than 160.
The search figures are truncation points, not rules: Google renders a title into roughly 580 pixels. X weights CJK and Hangul as 2 and counts every link as exactly 23 characters.
Frequently asked questions
Why is my paragraph count wrong?
A paragraph ends at a blank line, not every line break, so text with single breaks between paragraphs, which is what a PDF gives you, counts as one. Use Lines.
Does my text get uploaded anywhere?
No. Only the reading speed and common-words settings go in the URL, since a URL travels into browser history, a shared link and the next page's referrer header.
What counts as a sentence?
The same Unicode algorithm that finds word boundaries. It never breaks inside "3.5", and an abbreviation followed by a lowercase word or a digit, as in "e.g. this" or "Fig. 4", stays one sentence. A full stop followed by a capital letter does end one, so "Dr. Smith" counts as two sentences. Text full of abbreviated titles therefore counts high.
Is keyword density still worth looking at?
Not as a target. No density figure ranks a page. As a diagnostic: a term at 6 percent means repetitive writing, and a term you expected missing from the top ten means the page is about something else.