Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Content Marketing › Content Production

Writing for the Web

How people read on screens and how to write for it: front-loaded structure, headings and lists, plain language, evidence and examples, and accessibility.
Guidebeginner10 min readUpdated

How people read on screens

Nielsen Norman Group's research on how users read on the web, repeated across many studies since its first publication, reaches a consistent conclusion: most users scan a page rather than read it word by word, picking out words and sentences that look relevant and skipping the rest. Eye-tracking work shows attention concentrated at the top of the page and along the left edge, following headings, the first words of paragraphs and anything visually distinct, and tailing off as the page continues. Reading on a screen is also slower and more tiring than reading on paper, which lowers tolerance for text that does not get to the point.

The practical consequences the research draws are the basis of everything in this guide: highlight the words that carry meaning, write headings that say what the section contains, use lists where the content is a list, hold to one idea per paragraph, start with the conclusion and work down to the detail, and use roughly half the words a print version would. Readers who want more will read on; readers who want less get what they need and leave satisfied, which is the outcome Google's helpful content guidance describes and the one the E-E-A-T framing tries to capture.

Scanning behaviour also explains why click-through (CTR) from search and social depends so heavily on the first line a reader sees. The title and opening are read; the rest is scanned for reasons to stay. A page that buries its answer under three paragraphs of context has already lost the scanner, however good the answer is.

Structure: front-loading, headings, lists

Structure is where most of the gain is, because it is what the scanning reader uses. The three tools are the same at every scale: front-load the page, front-load each section and front-load each paragraph; make headings descriptive; and use lists where items are parallel.

ElementWhat scanning readers do with itRuleTest
Title and openingDecide whether to stayDeliver the answer or the promise in the first linesCould a reader stop after the opening and know the main point?
HeadingsNavigate; skip to the section they needEach heading describes its section specifically; a reader should know what is beneath it without reading onDo the headings alone tell the story of the page?
First sentence of each paragraphRead it; skip the rest if it does not applyCarry the paragraph's point in the first sentence; detail followsRead only first sentences: does the argument still hold?
ListsRead them fully, because they look quickUse for parallel items, steps and options; not for prose chopped into fragmentsIs every item the same kind of thing?
TablesScan for their row or columnUse when readers compare across two dimensionsCould the reader find their case in seconds?
Bold textStop on itReserve for lead-ins and the few phrases that must not be missedIs there little enough bold that it still stands out?
LinksNotice them; judge from the link textLink text says where the link goes; never "here" or "click"Would the link make sense read on its own?
Summary boxRead it instead of the pageThree to five sentences a scanner can act onDoes it stand in for the page?

Front-loading is the hardest habit for writers trained on essays, because the essay form builds to a conclusion. Web writing inverts it: conclusion, then support. Applied to the order of a whole message, whether a release, an email or a report, that sequence is the inverted pyramid structure. The structure in a brief, covered in the content brief tutorial, is the place to fix this before drafting, by putting the answer at the top of the outline. For search-led pages, the format and depth the results page rewards also shape the structure, which the search intent guide covers. Support articles add one more rule, writing for the question as customers phrase it, error messages included, which the help centre and documentation SEO guide covers.

Plain language and sentence length

Plain language is writing a reader can understand on the first pass. It is not simplified writing; it is writing without the inflation that forces readers to translate. The common offenders are abstract nouns where a verb would do ("make a decision" for "decide", "provide an explanation of" for "explain"), passive constructions that hide who does what, stacked qualifiers ("a somewhat more effective approach"), and jargon used because it sounds expert rather than because it is precise. Each is small; together they double the effort of reading.

Sentence length matters because long sentences hold several ideas in suspension and screens make that harder. The guidance is not "write short sentences" but "vary them": a long sentence that develops an idea, then a short one that lands it. A paragraph of uniformly short sentences reads as choppy; a paragraph of uniformly long ones reads as a wall. Read the draft aloud and the rhythm problems announce themselves.

Word choice follows the same principle. Use the specific word: not "utilise" but "use", not "solution" but the thing it is, not "content" when you mean "the guide". Use technical terms when they are the right terms and the reader knows them, and define them once when some readers will not. Spell out an abbreviation at first use. And cut the filler phrases that open paragraphs without saying anything, because the first sentence is the one scanners read.

Evidence, examples and specificity

The quickest way to tell a strong draft from a weak one is to count the adjectives that could be replaced by evidence. "A powerful approach" says nothing; "an approach the support team has used on every onboarding call since the spring" says something a reader can weigh. Specificity is what makes writing credible, and it is also what makes it useful, because the reader can act on a specific statement and cannot act on a vague one.

Evidence comes in a few forms, and each needs a source the reader could check: a figure with where it came from, a quote with who said it, a documented behaviour with the document, a first-hand observation labelled as one. The brief's proof list is where these are gathered; the writing stage is where they are placed so that each claim is followed closely by its support. Claims with no support are rewritten as clearly labelled opinion or cut.

Examples do the work explanation cannot. A worked example of a method, a before-and-after rewrite, a short scenario that shows the idea in use: each gives the reader a picture to check their understanding against. Label illustrative examples as such so readers know they are not case studies, and keep them concrete, with a named kind of organisation, a specific situation and a specific result. The E-E-A-T guide explains why search engines try to reward exactly this kind of first-hand, specific content; readers rewarded it first.

Illustrative example — not a real result

Rewriting an opening for scanning

A consultancy's guide on pricing a retainer opened with three paragraphs on the history of agency pricing before reaching its advice. Analytics showed the average reader left after 40 seconds, well before the first useful heading. The rewrite moved the advice to the opening: "Price a retainer on the outcomes it buys, not the hours it contains, and the three sections below show how to set the scope, the price and the review clause." Headings changed from "Background", "Our thinking" and "Approach" to "Define the scope in outcomes", "Set the price from value and capacity" and "Write the review clause". Each paragraph's first sentence now carried its point, two lists replaced a dense paragraph of options, and the word count fell from 1,900 to 1,150. Over the following quarter, median time on the page rose to just over 3 minutes and the share of readers reaching the enquiry form at the bottom went from 4% to 11%.

Accessibility in writing

Accessibility is often treated as a developer concern, but a large share of it sits in the writing. The Web Content Accessibility Guidelines from the World Wide Web Consortium describe content that is perceivable, operable, understandable and robust, and several of their success criteria are met or failed by the writer alone. Headings must be real headings, in order, with no levels skipped, because screen-reader users navigate by heading list and a page whose headings are bold paragraphs has no structure for them. Link text must make sense on its own, because links are often read as a list out of context; "download the template" works and "click here" does not. Images need alternative text that says what the image conveys, or is marked decorative when it conveys nothing. Abbreviations should be expanded at first use. Instructions should not rely on colour, position or shape alone ("the green button on the right").

The guidelines also address readability: WCAG includes guidance on making text understandable, and plain language, short sentences and defined terms serve it directly. Writing that scans well and writing that works with assistive technology turn out to be nearly the same writing, which is why the accessibility checks belong in the editing routine and not in a separate audit. Conformance requirements vary by jurisdiction and sector; treat the guidelines as the shared reference and check what applies locally.

Editing: a pass-by-pass routine

Editing everything at once catches little, because the eye cannot hold structure, sentence and word problems simultaneously. A pass-by-pass routine, each pass with one question, is faster and more reliable.

  1. The brief pass. Does the piece answer the brief's question, deliver its promise in the opening and follow its outline or explain the deviation? If not, fix this first; nothing below matters for a piece that answers the wrong question.
  2. The headings pass. Read only the headings. Do they tell the story of the page, in order, specifically? Rewrite any heading that would not make sense in a table of contents.
  3. The first-sentence pass. Read only the first sentence of each paragraph. Does the argument survive? Move the point to the front of any paragraph where it does not.
  4. The evidence pass. Mark every claim. Each has a source nearby, is labelled as opinion, or goes.
  5. The cutting pass. Remove preamble, repetition, hedges and filler openers. Most drafts lose a fifth of their length here without losing a point.
  6. The sentence pass. Read aloud. Break the sentences that run out of breath; join the ones that stutter; swap abstract nouns for verbs; cut qualifiers.
  7. The accessibility pass. Heading order, link text, alternative text, abbreviations expanded, no instructions by colour or position alone.
  8. The search and distribution pass. Title and description written for the reader who sees them in a results page or feed; internal links in place (SEO benefits from the same descriptive link text accessibility requires); the standalone sections the repurposing guide will need are marked.

The routine takes less time than it reads, because each pass is narrow. It is also teachable: a new editor given eight questions produces consistent work far sooner than one told to "tighten it up". The production hub places this routine at the edit stage, before expert review, so reviewers see a draft that has already been made readable.

Once the draft is ready, the SEO page publishing checklist covers the on-page checks that happen before it goes live.

Good web writing starts with a clear brief, and the content brief template helps you agree audience, purpose and structure with the writer before drafting begins.

If readers still leave early once a page scans well, the page on content getting no engagement looks at whether the topic, the channel format or the posting rhythm is the real cause.

Common mistakes

  • Opening with background. The answer arrives in paragraph four; the scanner left in paragraph one. Put the answer first and the background after, for those who want it.
  • Headings that label instead of describe. "Overview", "Approach" and "Conclusion" tell a scanner nothing. A heading should make sense in a table of contents on its own.
  • Lists made from chopped prose. Items that are not parallel, or that need the surrounding sentence to make sense, are prose in disguise and should go back to being prose.
  • Adjectives where evidence belongs. "Powerful", "seamless" and "robust" are places where a fact, a figure or an example was not gathered.
  • "Click here" and bare links. Link text that does not say where it goes fails scanners, screen-reader users and search engines in one move.
  • Editing in one pass. Structure, sentence and word problems hide from a reader looking for all three; one question per pass finds them.
  • Writing to a word count. Length comes from what the question needs. Padding to a target produces exactly the preamble the first pass removes.

Frequently asked questions

Does writing for scanning mean writing less?

It means writing so that the reader who scans gets the point and the reader who reads gets the depth. Length is set by what the question needs, not by a rule; a long piece with front-loaded paragraphs, descriptive headings and a summary up top serves both readers. What goes is the preamble, the throat-clearing and the paragraphs that exist to reach a word count.

Should every piece open with the answer, even a story-led one?

The opening should deliver what the reader came for, and for most informational content that is the answer or the promise. A story can open a piece when the story is the point or when the brief chose that angle deliberately, but it should be short and the reader should know within a few lines why they are reading it. On screens, patience is measured in seconds.

How does plain language square with writing for experts?

Experts prefer plain language too; they simply tolerate jargon better. Use the precise technical term when it is the correct one and the reader knows it, and explain it once if some readers will not. What plain language removes is the vague, inflated phrasing that experts and beginners both have to decode, not the vocabulary of the field.

Terms used on this page

Frameworks

Where this page is used

Part of these learning paths

Also related

Sources

  1. How Users Read on the Web — Nielsen Norman Group
  2. Web Content Accessibility Guidelines (WCAG) — W3C

Related