How Humans Read Websites (And Why They Mostly Don't)

Here is the statistic that quietly runs my entire career: in 1997, when Jakob Nielsen’s team put eye trackers on people and watched them use the web, they found that 79 percent of test users scanned any new page they came across. Only 16 percent read word by word.

That was 1997, when pages were slow and the internet was still a novelty you dressed up for. Three decades of infinite scroll later, nobody has become more patient. The modern estimate is worse: on an average visit, people read at most about 28 percent of the words on the page.

You’ve just spent about 40 seconds reading the paragraphs above. The rest of this page will probably get four seconds. Understanding what happens in those four seconds is the closest thing web design has to a unified theory of everything.

Nobody reads. Everybody scans.

When Nielsen Norman Group aggregated their eye-tracking studies, the heatmaps kept drawing the same shape: two horizontal stripes across the top of the page, then a vertical skim down the left edge. They called it the F-pattern, and it became the most misquoted diagram in our industry.

Here’s what the F actually is. The first line of text gets a full pass. The second gets a shorter one. By the third, the reader has stopped reading and started sampling, hunting the left edge for words that smell relevant. The F isn’t how humans read. The F is what the shape of giving up looks like.

The reframe

Visitors who scan aren’t being lazy. They’re triaging. The web is an all-you-can-eat buffet of pages promising to waste their time, and scanning is the plate-guarding technique that keeps the meal honest. Your page has to work as a menu before it can work as a meal.

Working memory is the real bottleneck

Scanning looks like a vision thing, but it’s a memory thing. Conscious attention can juggle roughly four chunks of information at once.

Your headline, your navigation, your hero image, that cookie banner, the thing they were trying to remember from the page before. Every element on screen draws from the same four slots.

A wall of unbroken text isn’t boring to your audience. It’s expensive. Reading a 400-word paragraph means holding hundreds of words in a four-slot workspace and assembling meaning on the fly, without dropping anything. Some visitors will do that work. Most will correctly conclude that your page is not going to pay them back for the effort.

Chunking helps precisely because of this. A tight list of five bullets isn’t five items in memory; it’s one chunk labeled “list of reasons,” with the details available on demand. You’re not dumbing it down. You’re moving the cost.

The F-pattern is a symptom, not a strategy

Somewhere along the way, “users read in an F-pattern” got mangled into “so put the important stuff in an F.” This is backwards. The F is what people do when a page gives them nothing worth a full pass. It’s the eye-tracking equivalent of sighing.

The fix isn’t to arrange your content into an F. It’s to build a page that keeps earning full lines: headings that make a promise, first sentences that pay it off, and paragraphs short enough that starting one feels cheap. Give the scanner a reason to become a reader, and the heatmap takes care of itself.

Eye-tracking in miniature: a full pass over the headline, a shorter pass over the subheading, then the skim down the left edge. The F is triage, not reading.

Show people a page with something that looks like an advertisement, and they will not see it, no matter what it says. Eye-tracking calls this banner blindness, and researchers have documented it consistently since 1998. The unsettling part: it fires on appearance, not content. Style a genuinely important announcement like a promo banner and watch it become invisible to everyone, including people actively looking for it.

The lesson goes both ways. Anything that looks like an ad is dead weight, even when it’s honest. And the fastest way to hide critical information on your own website is to give it promotional styling. That “Book a free consult” box you designed to look like a banner? To your visitors’ brains, it’s a territory they’ve already agreed to ignore.

Information scent: why they click

The best model for navigation on the web comes from a 1999 paper with the best name in the field: information foraging. Peter Pirolli and Stuart Card compared people to animals hunting for food, estimating the value of a patch before committing energy to it. On the web, links are the prey, and information scent is how strong the words on a link smell like the thing the visitor wants.

A link that says “learn more” has no scent. It could lead anywhere, which means it leads nowhere. A link that says “see our pricing breakdown for a five-page site” smells exactly like the thing a specific visitor came to find. Scent-rich links get clicked. Scent-free links get scanned past, and the visitor leaves to forage elsewhere, usually at a competitor’s patch.

The test

Read your page’s headings and links only, in order, and cover the rest. That skim is what four seconds of your visitor’s life actually experienced. If the story still makes sense, your page works. If it reads like a shoe of words fell down a staircase, now you know why nobody converts.

The rules that fall straight out of the science

None of this is opinion, which is my favorite kind of design advice. If people scan, if memory holds four chunks, if links need scent, then the checklist writes itself:

What the research says	What you do about it
Readers decide in seconds	Front-load the first two words of every heading
Four chunks of working memory	One idea per paragraph, three or four lines max
Bold and color direct attention	Emphasize sparingly; every extra bold trains the eye to trust none
Links need scent	Describe the destination, never “click here” or “learn more”
Long lines lose the eye	Keep text between 45 and 75 characters per line

That last one is a one-liner of CSS, which pleases me deeply. This is the entire measure you are reading right now:

.prose {
  max-inline-size: 66ch;
  line-height: 1.65;
}

Two declarations. No framework required.

But my audience is different

Every so often someone tells me their users are the exception: highly technical, deeply engaged, definitely reading every word. Sometimes that’s even true. Here’s the twist the research keeps confirming: plain, scannable writing helps experts too. Nielsen Norman’s plain-language studies found that even highly literate specialists complete tasks faster when the reading load drops. Experts don’t want your dense paragraphs. They want your dense paragraphs with the filler removed.

So the checklist above isn’t a simplification for dummies. It’s respect for the four slots everyone is working with, including the smart ones. Especially the smart ones. In my experience, the more senior the audience, the less patience they have for a page that wastes the slots.

Write for scanners. Readers take care of themselves.

This is the paradox I keep coming back to: the only reliable way to get someone to read your page is to design it for someone who won’t. Front-load the meaning. Break the ideas into chunks. Give every heading a job and every link a scent. Do it well and something lovely happens: the scanner slows down, the skim becomes a read, and the four seconds stretch into four minutes.

I keep most of these books on my library shelf, by the way. How to Read a Book is there, which is funny, because the web version of its core lesson is simply this: assume no one is reading, then make it impossible not to.

And if you looked at your own website while reading this and felt a small pity for its paragraphs, that feeling is data. Bring it to me.