The three things called “character count”
When a form says “maximum 160 characters,” it means one of three different measurements: graphemes (symbols a human sees), code points (Unicode’s units, what most programming languages count), or bytes (storage, dependent on encoding). For plain English they coincide; add emojis, accented letters, or non-Latin scripts and they diverge fast — café is 4 graphemes but can be 5 code points, and Turkish or Arabic text doubles or more in UTF-8 bytes. SMS is the classic trap: a single non-GSM character switches the whole message to UCS-2, dropping the per-segment limit from 160 to 70.
Practical length targets for the web
| Surface | Sweet spot |
|---|---|
| SEO title tag | 50–60 characters (pixel-truncated ~580px) |
| Meta description | 150–160 characters |
| Headline | 6–12 words |
| Sentence (readability) | 15–20 words average |
| Paragraph (web) | 40–80 words, 2–4 sentences |
These are starting points, not laws — but exceeding the title and description ranges reliably gets your text truncated with an ellipsis in search results, which costs clicks.
What reading time actually estimates
The reading-time figure divides word count by an average adult silent-reading rate (~225 words per minute). It estimates continuous prose reading: technical content with code, tables, or dense terminology reads 2–3× slower, and skimming reads much faster. Use it comparatively — to balance article lengths or chapter sizes — rather than as a promise to readers. For spoken scripts (presentations, voiceovers), use ~140 wpm instead; speech runs much slower than reading.
Counting across languages
Word counting assumes spaces separate words — which fails for Chinese, Japanese, and Thai, where segmentation requires dictionaries. Character counts remain meaningful for those languages (and are the standard billing unit for translation in CJK). For agglutinative languages like Turkish or Finnish, word counts run lower than English for the same content, since one word carries what English spreads across several — another reason character-based limits travel better than word-based ones.