How visitors actually read — and what that means for your site

There is one number that changes work on website texts as soon as you have heard it once: on an average visit, about 25 percent of the words are read.

Not 25 percent of the pages. 25 percent of the words on the page that is actually looked at.

Whoever takes that seriously writes differently. And above all: they build pages differently.

Three ways of reading

The visitor decides within milliseconds what to do. After that they switch into one of three modes — or they leave.

Skimming

The first rough pass. Attention is a large, blurry lens. It takes in a lot at once, but nothing precisely.

The purpose is orientation: does this fit my problem? Am I in the right place? This question has to be answered quickly and cheaply, before the visitor decides whether a closer look is worth the effort.

What is perceived in the process is well researched — through eye tracking. On the first pass, the eyes concentrate on:

• images
• headings
• text passages highlighted in colour
• elements on the left edge

Text is not read. Headings are scanned for keywords.

Scanning

The mixed mode, and the normal case for experienced internet users. The most informative elements are focused on:

• headings
• sentences with highlighted keywords
• the starts of paragraphs
• conspicuous elements: lists, tables, infographics
• striking images or info boxes

The pattern here is the F-pattern: the gaze runs across at the top, then across more briefly, then down the left edge.

Reading

The concentrated mode. Sequential, in detail, deeply processed. It occurs much less often — as a rule visitors read only what they rated as highly relevant while skimming.

(Rheinwerk, Website-Konzeption und Relaunch, ch. 2.2)

The blur test

From this insight follows a test you can do in thirty seconds and that shows a surprising amount:

Squint until the page becomes blurry. Or put a blur filter over a screenshot.

Then look at what can still be made out. The elements meant to be perceived while skimming — headings, important images, the call to action — must still stand out when blurred.

Two findings this test regularly produces:

Headings too small or too similar to the running text. They disappear in the blur. For the skimmer they therefore do not exist.

Elements that draw attention but contribute nothing. The classic is the oversized decorative image that is the only thing you can see when blurred — and that carries no information.

And a second principle from the same observation: same function, same form. All headings of one level look the same, all buttons look the same, all links look the same. Then the visitor recognises them as a group and skims faster. If every page brings its own design, they have to learn anew each time.

(Rheinwerk, Website-Konzeption und Relaunch, ch. 2.2)

What follows for the structure of a page

Key messages at the start and at the end. What matters belongs at the top as a brief summary — and once more at the end. The skimmer sees both.

The most important point of a paragraph first. Because the starts of paragraphs are read and the rest often is not. Whoever puts the key sentence at the end of the paragraph hides it.

That is the reverse of what you learn at school — there you build up and reach the conclusion at the end. On the web, the conclusion is the beginning.

One topic per paragraph. Not three. A paragraph that deals with three things is found for none of the three.

Meaningful subheadings. They are the backbone for both modes: the skimmer uses them as signposts, the reader as an outline. A subheading that does not say what is in the following section is wasted.

Use lists, tables and infographics. They are explicitly targeted while scanning — users have learned that important information is there and not in the running text.

Do not overload. The core content has to be graspable at first glance.

Use familiar elements. This is the most underestimated point: if your website uses controls the way the visitor knows them from other websites, they save the mental effort it takes to learn them anew — and have that capacity free for your content. An original menu is a tax on your visitors' attention.

And the connection to findability: the F-pattern explains why important terms belong at the front of the page title and the description. The same gaze pattern runs on the search results page — the heading of the result, then the description below it, the rest is only grazed.

(Rheinwerk, Website-Konzeption und Relaunch, ch. 2.2)

Why anyone comes at all

A step back, because it puts the above in context: every website visit goes back to a need.

The typical patterns:

Need	Example action
Information, thirst for knowledge	Researching how something works
Advantage, a good deal	Comparing prices
Entertainment, distraction	Watching a video
Belonging, contact	Exchanging views
Convenience, saving time	Ordering online instead of going there
Recognition, reputation	Showing something that impresses
Getting or giving help	Describing a problem

Most actions have several motives. Whoever asks for a quote may at the same time want to save time, make sure they do nothing wrong, and get a good price.

The course of the decision is similar for everyone:

1. Motivated search — someone has a problem and wants to solve it.
2. Arrival and relevance check — within milliseconds it is judged whether the page fits the problem.
3. Comparison with one's own requirements — once the first check is passed, it is checked whether the page meets the concrete expectations.
4. Decision.
5. Action — enquiry, call, purchase.

The point where most websites lose is step 2. Not because the offer does not fit, but because in the first seconds it is not recognisable that it fits.

(Rheinwerk, Website-Konzeption und Relaunch, ch. 2.1)

Meet expectations — or lose visitors

Over the years visitors develop an intuition for how websites work. These expectations can be met or deliberately broken. What you should not do is break them by accident.

The continuation expectation

A visitor expects a link to lead to what its link text announces.

Everyone knows what is behind “Basket”. Behind “Interesting” or “Worth knowing”, nobody does.

The worse case: you click “Download” and end up, unexplained, in a registration form. The visitor then first has to notice that they were led the wrong way, find out where they are, and reorient themselves. That costs attention and goodwill.

The test for this is simple and effective: have someone who does not work in your business go through your menu items and buttons and say for each one what they expect behind it. Where several people think for a long time or guess wrong, the label is the problem.

And that does not only apply to a new build — for an existing website it is worth checking regularly.

The process expectation

With multi-step processes — ordering, booking an appointment, a multi-page form — the visitor plans in their head what is still to come. If reality deviates from that, they are irritated.

Uncertainty about the process is one of the most common reasons visitors abandon.

What helps:

• a label for the current step
• a progress indicator — how many steps are there in total, where am I
• a note on how to go back and what comes next
• clarity about the consequences: is the click on “Next” binding, or can I still change things afterwards?

The classic failure: a booking process where you cannot see how many steps are coming, why the “Next” button is greyed out, and what would have to be done for it to continue.

The guiding question after every building step is: can the visitor say at any time where they are, where they came from and what is still to come?

Feedback

Visitors expect a reaction at certain points. This shows most clearly with forms.

If the feedback after submitting is missing or incomprehensible, the visitor does not know whether it worked. A common consequence: they submit again — and you get the same enquiry three times.

Worse is a missing or cryptic error message. The classic mistake is a single reproachful line (“Invalid input!”) above the whole form, without saying which field is meant.

The right way is: understandable feedback directly at the faulty field, saying exactly what is wrong and what to do.

(Rheinwerk, Website-Konzeption und Relaunch, ch. 2.7)

Seven is the upper limit

A value that runs through the whole design: working memory holds about seven units.

In practice that means: seven is a sensible upper limit for menu items, for bullet points in a list, for sections on a page. Beyond that, something falls out at the back.

That is no law of nature, but a good rule against the tendency to fit everything in. A menu with fourteen items is not a menu but a list.

Related to this is decision fatigue: every decision costs something. Whoever makes the visitor choose in too many places exhausts them before they reach the actual decision. For choices, a few clearly distinguishable options help. For pros-and-cons decisions, about four to five concise arguments — not twelve.

(Rheinwerk, Website-Konzeption und Relaunch, ch. 2)

How to check this on your site

The five-second test. Show someone your home page for five seconds, then take it away. Ask: what does this business do? For whom? What am I supposed to do here? If the answers do not come, the top area is not working.

The blur test. Squint. What stands out? Is it the right thing?

The paragraph-start test. Read only the first line of every paragraph. Does that give you the content of the page?

The heading test. Read only the headings. The same.

The link label test. Have someone guess what is behind each menu item.

The form test. Fill in your contact form wrongly on purpose. Does it say which field is wrong and why? And on correct submission: is it clear that it worked?

What to do

1. Rebuild the top area of the home page around the millisecond question: does this fit my problem? Who are you, what do you do, for whom.

2. Rewrite the subheadings. They carry most of the load.

3. Turn the paragraphs around. Key message first.

4. Put summaries at the start and end of longer pages.

5. Check the menu labels with real people, not in your head.

6. Add progress indicators to multi-step processes.

7. And look at your error messages. They are the place where visitors give up, and the least often revised text on a website.

What I tell clients about this: most website texts are not badly written but wrongly arranged. The same content, rearranged and given load-bearing subheadings, works differently — because for the first time it fits the way people actually read.

Sources

• Rheinwerk, Website-Konzeption und Relaunch, ch. 2.1 (why users use websites) — that every online action is based on a need and actions can have several motives, the complete table of typical needs with example actions, and the five-step course of decision from the motivated search through the relevance check in milliseconds and the comparison with one's own requirements to decision and action.
• Rheinwerk, Website-Konzeption und Relaunch, ch. 2.2 (floating and focusing attention) — the distinction between skimming, scanning and reading with their properties and the comparison table, the eye-tracking findings on images, headings, text highlighted in colour and elements on the left edge, the share of about 25 percent of words read, the F-pattern including its effect on search result pages and the consequence for page title and description, the list of elements focused on while scanning, the blur test with the two typical findings on headings that are too small and oversized decorative images, the principle of the same form for the same function, and the four consequences for the concept (lean skimmable design, structure for scanning with summaries at the start and end and the most important point at the start of the paragraph, good texts, familiar standard elements).
• Rheinwerk, Website-Konzeption und Relaunch, ch. 2.7 (visitors' process expectations) — the continuation expectation with the basket example and the case of the misleading download button, the test by asking outsiders about link labels and the recommendation of regular checks, the process expectation with the finding that uncertainty about the process is one of the most common reasons for abandoning, the four remedies (step label, progress indicator, notes on the way back and the next step, clarity about how binding it is), the booking example with the greyed-out Next button, the guiding question about location, origin and next step, and the requirements for feedback and for error messages directly at the faulty field.
• Rheinwerk, Website-Konzeption und Relaunch, ch. 2 (key messages) — the brain's principle of economy and cognitive ease, the stages of memory and intake in units with the upper limit of about seven for navigation items, list items and sections, and decision fatigue with the recommendations of few distinguishable options and four to five arguments for pros-and-cons decisions.
• Own audit practice — the five-second test, the paragraph-start and heading tests, the observation that missing submission confirmations lead to multiple enquiries, the classification that most website texts are not badly written but wrongly arranged, and the order of the work steps.