How reading time is calculated
A short reference for writers and editors: where the "5 min read" estimates come from, what the words-per-minute numbers mean, and why the same piece takes three times longer to read aloud.
The basic formula
Almost every reading-time estimate you see online — the "5 min read" badge on Medium, the time pill on Pocket and Instapaper, the small print at the top of news articles — uses the same one-line formula:
reading time = word count ÷ words per minute
The word count comes from counting whitespace-separated tokens in the article body (excluding navigation, ads and footers). The words-per-minute number is the assumption — and that's where the differences between estimates come from.
The words-per-minute numbers people actually use
| Mode | Reader | Words / min |
|---|---|---|
| Reading silently | Slow / studious | 100–150 |
| Reading silently | Average adult, prose | 200–250 |
| Reading silently | Fast, skimming for gist | 400–500 |
| Reading silently | Speed-reading (trained) | 700–1,000+ |
| Reading aloud | Audiobook narrator | 150–160 |
| Reading aloud | News broadcaster | 150–170 |
| Reading aloud | Conversational speech | 130–150 |
| Reading aloud | Auctioneer / sports commentary | 250–400 |
The most common number for online articles is 238 words per minute. This comes from a 2019 meta-analysis (Brysbaert) reviewing 190 reading-rate studies — currently the most cited figure in the academic literature. Older estimates often used 200 wpm as a round number; newer tools tend to use 250.
Different platforms pick different defaults:
- Medium uses an internal estimate that lands at roughly 265 wpm.
- Pocket aims for around 220 wpm.
- Instapaper publishes its formula as 250 wpm.
- The WordPress reading-time plugin defaults to 200 wpm but is user-configurable.
- This site (count.) uses 230 wpm — close to the academic centre, slightly conservative.
Why the estimate is always a lie (in a useful way)
Reading speed varies enormously between readers, between days, and even between paragraphs of the same article. The "5 min read" badge isn't a measurement — it's a comparison ranking. Its job is to let you tell at a glance which article in your feed is short, medium or long, not to predict the actual seconds you'll spend.
Three factors that the simple formula doesn't catch:
- Content type. A page of dialogue is read 30–50% faster than a page of dense expository prose. Code samples and tables drag the rate down a lot — readers pause on each line. Bullet lists are read faster than the equivalent paragraph.
- Format. Long lines (wider than ~75 characters) slow readers because their eyes lose place between lines. Justified text with poor hyphenation does the same. Black text on a clean light background is fastest; low-contrast or fancy fonts are slower.
- Language. Reading rates are typically reported for English. Languages with longer words (German), denser scripts (Mandarin, Japanese), or right-to-left layouts (Arabic, Hebrew) all change the effective number — though the words-per-minute count is then misleading because the unit "word" is different too.
Speaking time versus reading time
If you're writing a script, a podcast intro, or a speech, the reading-time estimate is the wrong number. Switch to speaking time, which is roughly 140–150 words per minute for natural conversational delivery.
| Total words | Reading silently (230 wpm) | Speaking aloud (145 wpm) |
|---|---|---|
| 200 | 52 s | 1 min 23 s |
| 500 | 2 min 11 s | 3 min 27 s |
| 1,000 | 4 min 21 s | 6 min 54 s |
| 2,000 | 8 min 41 s | 13 min 48 s |
| 5,000 | 21 min 44 s | 34 min 29 s |
Rule of thumb for scripted speaking: target around 150 words per minute on the page. A 15-minute talk needs roughly 2,250 words of prepared text. A 3-minute video script wants about 450 words.
How count. shows reading time
The reading-time block in count. shows two numbers: silent reading at 230 wpm, and speaking time at 145 wpm. They're both rounded to the nearest whole minute above 30 seconds (so a 1,180-word article reads "5 min" not "5 min 8 s"), with the seconds spelled out for shorter texts.
The count is taken on the full text as you paste or type — there's no "exclude code blocks" or "exclude quotes" because we don't have a way to detect those reliably in arbitrary input. If you're estimating reading time for an article that's mostly code, divide the resulting estimate by about two to be safer.
Common questions
What WPM should I use for a blog post?
Use the platform's number if there is one (Medium, WordPress). Otherwise 230–250 wpm is a defensible middle ground for general adult readership. Lower it (180–200) for technical writing with dense vocabulary; raise it (300+) if your audience is dedicated and the prose is easy.
Why are speed-readers so much faster?
Most "speed reading" gains come from suppressing subvocalisation (the inner voice that "speaks" each word) and from learning to recognise common phrases as single units. The trade-off is comprehension on dense material — speed-readers do well on light prose and poorly on dense argument-heavy texts. For most articles the comprehension penalty isn't worth the time saved.
Is there a reliable way to measure my own WPM?
Pick a familiar 1,000-word piece you've never read carefully. Read it once. Note the time. Take a short quiz on what it said (a friend can write three questions, or a tool can generate them). If you got 80%+ right, your effective WPM is 1,000 ÷ minutes-elapsed. If you got less, slow down and try a different piece. Calibrating once a year is plenty.
Why do news articles seem to take less time than the estimate?
Two reasons. First, you're probably skimming, not reading every word — average news-consumption rates are closer to 400 wpm than 230. Second, news writing is deliberately structured so that you can stop after the first paragraph and still know the story. Both push your actual time well below the "X min read" badge.
How accurate is the estimate for non-English text?
Less accurate than English. Word counts depend heavily on tokenisation, and tokenisation is a notional concept in languages like Japanese or Chinese where there are no spaces between "words". For those, a character-count-divided-by-character-per-minute rate is more reliable than a word count — but most online reading-time tools don't do that.