Sometimes, and that is usually bad news. The F-pattern is not a reading style people prefer. It is what scanning looks like when a page gives them nothing better to work with. Eye-tracking shows it happening on walls of text with no headings, no bolding, and nothing to grab.
This is the part that gets lost when the F-pattern gets passed around as a design rule. Designers hear "people read in an F" and start arranging layouts to fit the shape. That is backwards. The F is a symptom, not a target.
The Z-pattern is a separate story, and a weaker one. It circulates with the same confidence and far less evidence behind it. Both deserve a closer look, because the way most teams apply them does not match what the research actually found.
It is a scanning pattern where the eye moves across the top of the content, then across again slightly lower, then down the left side. Nielsen Norman Group documented it with eye-tracking, describing how "in left-to-right languages, text on the left and towards the top of the page is read more than text on the right or towards the bottom of the page".
The evidence goes back a long way. NN/g cites an early 2000s study with more than 45 participants and more recent work with 47 people, and the pattern shows up consistently across them. It also appears on mobile and in right-to-left languages, mirrored.
One finding is worth writing on a wall. NN/g states that "first lines of text on a page receive more gazes than subsequent lines of text on the same page". Whatever else you take from this research, the top of a block of text is the most valuable real estate you have, and most pages waste it on a warm-up sentence.
No. NN/g titles its own article on the subject "F-Shaped Pattern of Reading: Misunderstood, But Still Relevant", and the misunderstanding is exactly this. The F is what happens when a page fails to guide someone. When it appears in your session recordings, the layout is at fault, not the reader.
The research is specific about the conditions that produce it. NN/g describes three things happening together: the text lacks web formatting such as bolding, bullets, or subheadings, the user is prioritising efficiency, and the user has no strong commitment to the content. Two of those three are about your page, not your visitor.
So the goal is not to design an F-shaped layout. It is to make the F unnecessary. Give people headings that say what each section contains and they stop sweeping blindly and start jumping to what they need.
We have watched this play out in redesigns for years. Break a dense page into clearly labelled sections and the whole reading behaviour changes. The content did not get shorter. It got navigable.
The Z-pattern is the idea that on a sparse page the eye travels across the top, diagonally down to the bottom left, then across the bottom, so you should place a logo, a headline, and a call to action along that path. It is a widely repeated design heuristic. It does not rest on the same evidence base as the F-pattern.
Here is what we can say with confidence. NN/g published research identifying four text scanning patterns from eye-tracking: the F-pattern, the spotted pattern, the layer-cake pattern, and the commitment pattern. The Z-pattern is not among them. That is not proof it never happens, but it does mean the most cited eye-tracking body of work in the field did not name it as one of the patterns it observed.
Our honest read is that the Z-pattern describes a reasonable layout convention rather than a measured behaviour. Putting your logo top left, your navigation top right, your headline in the middle, and your button below does produce good pages. That works because it matches convention and hierarchy, not because eyes trace a Z.
The distinction matters when someone uses the Z to justify a decision. If a designer argues a button belongs bottom right because the Z ends there, they are appealing to an authority that is thinner than it sounds. Argue from hierarchy and contrast instead. Those you can actually demonstrate.
Three more, and two of them are more useful to design for than the F. The layer-cake pattern, the spotted pattern, and the commitment pattern each describe a different reader with a different goal, and a good page supports all of them at once.
The layer-cake pattern is the one to design toward. NN/g defines it as a pattern that "consists of fixations placed mostly on the page's headings and subheadings", with readers dropping into body text only once a heading tells them they are in the right place. That is a reader being served well, and it is why question-shaped headings do so much work.
The spotted pattern is people hunting. NN/g describes readers focusing on words that "visually stand out in the text because they are styled differently (e.g., links, differently colored words, bolded words, bulleted lists)". This is the reader looking for a price, a phone number, or one specific term. Bold the things people hunt for and you serve them in seconds.
The commitment pattern is the rare one where "users fixate on all or most content words in the text passage". It happens when someone genuinely cares. You cannot design your way into it, but you can avoid designing your way out of it with cramped line lengths and low contrast.
Make the page scannable at three levels: headings, emphasised phrases, and the first line of every block. NN/g recommends chunking content into sections, using meaningful subheadings, applying visual styling to keywords, and naming links well. That list has not changed in years because it keeps working.
Meaningful is the operative word for headings. "Our approach" tells a scanner nothing. "How long a build takes" tells them exactly whether to stop. If your headings only make sense to someone who has read the section beneath them, they are labels rather than signposts and the layer-cake reader cannot use them.
Front-load every paragraph. Given the finding that first lines get more attention than the lines after them, a paragraph that opens with context and lands its point at the end is spending its best asset on throat-clearing. Put the claim first and the support after.
Use emphasis sparingly enough that it means something. If a page bolds twenty phrases, none of them stand out and the spotted-pattern reader gets no help at all. We treat bold as a budget rather than a style. This all connects to the wider structure of a page, which we covered in our piece on what visual hierarchy is and why it matters.
Yes, and NN/g documents mobile scanning in the same research. The shape compresses because the column is narrow, but the underlying behaviour holds. The top of the visible area and the start of each text block still get disproportionate attention.
What changes on mobile is how quickly the top of the page disappears. On a phone the visitor sees a fraction of what a desktop visitor sees, so the first block carries even more weight. A hero that spends its height on a decorative image pushes your only guaranteed-seen text off the screen.
Narrow columns also make headings more important, not less. With less text visible at once, a reader has fewer landmarks and needs clearer ones. We have found that mobile-first layouts which get headings right tend to read better on desktop too, which is a good sign the hierarchy is genuinely working rather than being propped up by screen size.
The related question of what belongs in that first screen is worth its own look, and we went through it in our article on whether above the fold still matters.
No. Left-weighting is a real effect in left-to-right languages, but treating it as a rule produces lopsided, boring layouts. The finding is that unstyled text on the right gets less attention, not that anything on the right is invisible.
Contrast beats position. A high-contrast button on the right will out-pull a low-contrast line of text on the left every time. The scanning research describes default behaviour on undifferentiated content. The moment you introduce visual weight, you are changing where the eye goes, which is the entire point of design.
Where left-alignment genuinely matters is body copy. Centred paragraphs force the reader to find a new starting point on every line, which fights directly against the way scanning works. Centre a headline if you like. Do not centre a paragraph of any length.
It also matters for scannable lists of information, where a ragged left edge slows everything down. Keep the left edge clean and predictable and let the design create interest elsewhere.
Use session recordings and heatmaps rather than assuming. Tools like Hotjar and Microsoft Clarity show where people click, how far they scroll, and where they hesitate. That is not eye-tracking, but it answers the practical question of whether your page is being used the way you designed it.
The cheapest test needs no tools at all. Show someone the page for five seconds, take it away, and ask what it was about and what they could do next. If they cannot answer, no scanning pattern is going to save the page. We run this on every project and it finds problems that hours of internal discussion do not.
Scroll depth is the other number worth watching. If most visitors never reach a section, its position is the problem rather than its content, and moving it up is a faster fix than rewriting it. Long-form pages especially benefit from this check, and we went through the reading experience in detail in our guide to designing a blog post page people actually read.
Rewrite your headings so each one says what its section contains, then rewrite the first sentence of every paragraph so it leads with the point. Those two changes address the layer-cake reader and the F-pattern reader at the same time, and neither requires a redesign.
After that, look at what you have emphasised. Cut the bolding down to the handful of phrases a hunting reader would actually be looking for. If everything is emphasised, nothing is, and you have removed the one tool that helps a spotted-pattern reader.
If you have a page that is well written and still not landing, that is usually a scanning problem rather than a copy problem, and it is a satisfying one to fix. Send it over at phoenix.studio and we will tell you what we would change first.
Tell us where you want to go. We'll tell you how we'd get you there.