How accurate is the reading time estimate shown on blog posts?

Most reading-time widgets assume a fixed 200 WPM. Brysbaert's 2019 meta-analysis put the true value between 238 WPM (non-fiction) and 260 WPM (fiction). Medium uses 265 WPM. On top of that, users typically read only 20-28% of the words on a page. The displayed time is a theoretical ceiling; actual dwell as a percentage is much lower.

Is GA4's average engagement time the same as reading time?

No. GA4 only measures the time the tab is in focus. It does not check for active reading, scrolling, or interaction. Plausible's 28-day comparison showed GA4 underreports real engagement time by 54.7% on average and up to 80% on some pages. The cause is non-engaged sessions inflating the denominator.

How is the reading ratio calculated?

Expected reading time = (word count / 238) * 60 * 1000 milliseconds. Actual active time accumulates only while document.visibilityState is 'visible'. Reading ratio is the ratio of these two. Below 0.2 is skim, 0.2-0.6 is scan, 0.6-1.2 is real reading, above 1.5 is either deep reading or AFK.

If reading ratio exceeds 1.0, was the content really read carefully?

Not on its own. A ratio above 1.0 means the reader stayed longer than expected. That could be deep reading, or it could be a tab left open. Combine with scrollTracker's micro movement ratio, direction changes count, and temporal pattern. High direction changes plus micro movement means real reading. No movement means AFK.

Should TR and EN content use the same WPM baseline?

No. Second-language readers are roughly 30% slower. For a TR primary post being read by an EN reader, baseline should be around 165 WPM instead of 238. The next step is segmenting reading ratio by GA4's language dimension. A future post will look at this BigQuery analysis on ceaksan.com's own data.

Does light vs dark mode affect reading time?

Academic data favors light mode. Buchner and Piepenbrock's positive polarity advantage studies consistently found light mode outperforms dark mode on proofreading. A 2025 MDPI tablet study measured equal reading speed (243-249 WPM) but users perceived dark mode as less tiring. I will cover this in detail in a follow-up post.

Reading Time, Recalculated: 238 WPM, the 20% Rule, and Scroll Signals

The reading time estimate at the top of blog posts usually assumes 200 WPM. Brysbaert's 2019 meta-analysis showed it is actually 238 WPM. NN/g eye-tracking data says users only read 20-28% of words on a page. GA4 underreports actual engagement time by an average of 54.7%. A reading ratio metric that unifies these three findings, plus scrollTracker integration.

TL;DR

The '5-minute read' label on blog posts almost always assumes 200 WPM. Brysbaert's 190-study meta-analysis found 238 WPM for non-fiction and 260 for fiction. NN/g eye-tracking shows people read only 20-28 percent of the words on a web page. GA4 underreports that same time by 54.7-80 percent. So the estimate and the measurement drift in the same wrong direction.

View Subscription Sign In