Section 1
What the eyetracking actually shows
Jakob Nielsen's foundational study reported that 79% of test users scanned any new page they encountered; only 16% read word-by-word. The 2006 eyetracking work made the behavior visible: on text-heavy pages, heatmaps showed two horizontal stripes across the upper content area, then a slower vertical stripe down the left side, the F-shape. Nielsen Norman Group's follow-up research eleven years later confirmed the pattern persists, including on mobile, and clarified what it is: not a design goal but a symptom of users economizing effort on pages that give them no better guide. The F-pattern emerges when formatting fails, when nothing on the page signals where the valuable words are, the eye defaults to where valuable words usually live: the top, the start of lines, the left edge. Everything else gets peripheral vision at best. Your visitors are not lazy readers. They are efficient foragers on pages that hide the food. For the step that usually comes next, see [The Psychology of Website Conversion: Why Buyers Say Yes (or Quietly Leave)](/blog/psychology-of-website-conversion).
Section 2
The scanning patterns, and what each one costs you
The F-shape is the famous pattern, but NN/g's text-scanning research catalogs several, and each punishes a different design mistake. The table below summarizes the major patterns, when they appear, and what gets missed, which is the commercially relevant column, because what gets missed on a service site is usually the differentiator, the proof, or the call to action. Two implications deserve emphasis. First, the F-pattern's decay means position is an argument: content earns attention by where it sits, not by how good it is. Second, the patterns are triggered by formatting, which means you choose the pattern your visitors use. Strong subheadings, bold keywords, and front-loaded sentences shift users from the lossy F-scan toward the layer-cake pattern, scanning headings and deciding deliberately where to commit reading. That shift is free conversion: same words, repositioned.
Section 3
Writing and layout for scanners
Designing for scanners is mostly a writing discipline. Headings must carry the argument by themselves, a visitor who reads only your subheadings should still get the pitch, which is the layer-cake pattern working for you. Front-load every heading and paragraph: NN/g's guidance is that the first two words of a line do disproportionate work, because the F-scan's vertical stripe samples exactly those words. Use the inverted pyramid, conclusion first, support after, at page level and paragraph level. Bold the facts a spotted-pattern hunter seeks: prices, numbers, deadlines, outcomes. Keep paragraphs to one idea, lists short, and line openings varied so the bypassing pattern does not swallow your bullets. Nielsen's original study found that concise, scannable, objective rewrites improved measured usability dramatically, not because the words got prettier, but because the page stopped fighting how eyes actually move. A useful companion to this piece is [The Economics of Page Speed: What a Slow Website Actually Costs You](/blog/page-speed-economics).
Section 4
Scanning and the placement of conversion elements
The conversion consequence of scanning is positional: calls to action, proof, and differentiators must sit on the scan path, not merely on the page. Practically, that means your primary CTA appears high and left-anchored or full-width, never only at the bottom of a long page the F-scan will not reach; proof elements sit adjacent to the claims they support, where the eye actually pauses; and key differentiators live in headings, not in body copy that exists mainly as texture. It also means auditing with scan logic: read only your headings, bolds, and buttons top to bottom, is the case made? Inside ConvertOS, the web-design module of our LeverageOS framework, that headline-only readthrough is a standard QA step on every page template. If you want your own site audited the same way, that is a strategy call, and it takes about an hour to find what scanners never see. If you are turning this into practice, [AI Chat for Lead Capture: Turning Website Visitors Into Booked Calls](/blog/ai-chat-lead-capture-booked-calls) maps the adjacent system.