Users scan documentation, they don't read it

What it is

“Users scan, they do not read” is shorthand for common documentation tasks, not a claim that everyone always avoids continuous reading. Task-conditioned documentation reading modes often form a locate–scan–read sequence: a person uses search, a table of contents, or find-in-page to reach candidates, judges relevance quickly, and then reads the matching procedure, explanation, or constraint closely. Learning an unfamiliar domain, reviewing high-stakes requirements, and following a complex argument may instead invite sequential reading.

Scannability therefore does not mean reducing prose to fragments. It means helping readers enter at the right place, recognize a match, and obtain sustained explanation when needed. One person can switch modes as the task changes.

Why it happens

Information goal, urgency, prior knowledge, genre, device, and entry route all shape reading. A troubleshooter arrives with an error or symptom to match; an operator needs one step or parameter; a learner builds conceptual dependencies; an auditor may inspect every condition. Search and deep links also land people in the middle, so the opening is no longer a shared starting point. Scanning narrows the candidate set, while close reading carries reasoning, conditions, and risk. The two behaviors are complementary.

Small mobile viewports increase scrolling and the need for clear location and short routes. Screen-reader users scan nonvisually through heading lists, landmarks, links, and find commands. A design optimized only for visual skimming collapses into a long undifferentiated stream when semantic structure is missing. When a search snippet promises something different from its landing section, readers must return and locate again.

Studying it

Design studies by task rather than demographic label: retrieve a fact, perform an operation, diagnose a fault, learn a concept, or verify a high-consequence condition. Capture queries, result choice, landing point, find-in-page use, heading jumps, rereading, answer accuracy, completion time, and continued search; ask when scanning became close reading. Scroll depth, dwell, and exit do not establish reading, comprehension, or satisfaction without answer and task evidence.

Test entry from the page top, internal search, external search, and deep links across desktop, narrow screens, keyboard, and screen reader. Check whether snippets predict their landing and whether query context survives. For assistive technology, observe whether the heading list locates the target and whether focus and reading begin there after navigation. Hold content and search ranking constant when comparing structure so retrieval quality is not mistaken for layout effect.

Where it stops holding

Sequential tutorials, policy or contract review, incident procedures, and conceptual arguments may require continuous reading. Critical prerequisites cannot be placed where selective readers can unknowingly skip them. Ordered content can still offer progress, step headings, and a way back to position; scannability and sequence are compatible. Release notes lean toward locating changes, whereas long-form explanation may prioritize argumentative continuity, so one density does not fit both.

Search and deep links aid location but do not guarantee correctness or understanding. High-stakes tasks need prerequisites and consequences to travel with the target passage, not just a fast hit. Multilingual, dynamic, and versioned documentation can change heading text or position, so stable landing behavior needs governance separate from the displayed title.

Applying it

• Inventory each document's primary tasks and expected modes: fact lookup, procedure, troubleshooting, learning, or audit. A page can provide a quick answer alongside optional rationale, constraints, and examples.
• Make search-result titles and snippets correspond to their sections, and land internal search, external search, and shared links on named subsections. Use stable anchor identifiers; preserve redirects when headings, locale, or versions change so old links do not fall back to the top or a wrong passage.
• On mobile, expose location, proportionate in-page navigation, and a route back to results without controls obscuring content. Build screen-reader navigation from semantic headings and landmarks. After a deep-link jump, move focus or reading position to the target heading while retaining necessary context.
• Validate the query–landing–answer–outcome chain: measure time to locate, answer accuracy, subsequent action, and repeated search for the same problem. A fast wrong answer is not a scanning success.

Related

• Same group: T3.01.2 Headings, lists, and code blocks carry the scan anchors · T3.01.3 One section answers one question
• Adjacent: T3.02.2 Help entries belong next to where problems occur · T3.02.3 Search must tolerate user vocabulary and synonyms
• Search terms: documentation scanning · information foraging · selective reading