How people use the web with assistive technology

A screen reader turns the page into speech, braille or both. On Windows it has two ways of working:

Mode	NVDA's name	What the keys do
Reading	browse mode	The arrow keys move a reading cursor through the page "as if it were a normal text document", and letter keys jump to the next heading, link or field
Typing	focus mode	"Almost all keys are passed to the control", so people can type into a field and use a widget's arrow keys

NVDA's user guide says it switches to focus mode by itself when people Tab to or click a control that needs it, and back again when they leave it. JAWS calls its typing mode Forms Mode. VoiceOver on a Mac has no modes: people move into an item to interact with it. ARIA's application role asks screen readers that intercept keys to pass them through to the page, and ARIA then requires that all the text and images inside it be reachable through a focusable element, because some users can reach only the focusable elements in it.

Most people don't read a long page from the top. WebAIM's survey #11 (2026) asked what respondents do first to find information on a lengthy page:

First step on a long page	Respondents
Move through the headings	67.8%
Use the Find feature	12.5%
Read through the page	9.7%
Move through the links	6.7%
Move through the landmarks	3.4%

28.3% use landmarks whenever they are there or often, and 88.3% find heading levels very or somewhat useful. Screen readers also gather these into lists: NVDA's Elements List shows the links, headings, form fields, buttons or landmarks on a page, and VoiceOver's rotor chooses which of them a swipe moves by. A link in such a list is read without the text around it, which is why eight links that all say "Details" can't be told apart, as "Read more" links that all say the same thing shows.

On a phone there are no keys. People drag a finger to hear what is under it, swipe right and left to move item by item, and double tap to activate. With VoiceOver they turn two fingers on the screen, the rotor, to pick headings or links, then swipe up or down. TalkBack has reading controls for the same job and a TalkBack menu, opened with a three-finger tap from version 9.1. The keys and gestures for each task, in each screen reader, are on screen reader shortcuts by task.

A refreshable braille display shows a line of braille, usually 40 to 80 characters, by raising and lowering pins. NVDA's user guide and Apple's iPhone guide both cover braille displays. People who are deafblind may read the whole page this way. W3C's resource adds that not all people who are blind know braille.

A screen magnifier enlarges part of the screen, so a person sees a small window onto the page and moves it around. The Understanding document for 1.4.13 says the magnified view "generally follows the mouse pointer or keyboard focus", and iPhone's Zoom has a Follow Focus setting that tracks the selection, the insertion point and typing. Browser zoom is the other way to enlarge: at 400% in a 1,280-pixel window the page has 320 CSS pixels to fit in, as pages that scroll sideways on a narrow screen explains.

People often combine magnification with speech. W3C's resource describes people with low vision who magnify the screen, change the text and background colours, and "use text-to-speech to hear the text while they are reading". Windows Magnifier can read text aloud, and Apple's guide says iPhone's Zoom can be used together with VoiceOver.

Because only part of the page is in view, anything that appears on hover has to stay put. The Understanding document says that when added content is large, a magnifier user may need to pan to see it, "which is impossible unless the user is able to move their pointer off the trigger without the additional content disappearing", and that content covering the page must be dismissible without moving the pointer or focus. Tooltips that vanish on approach and can't be dismissed is that failure.

Reading support changes how text is presented, not how much of it is on screen. W3C's resource lists these among reading assistants:

• text-to-speech that reads the content aloud and highlights the text as it is read;
• changing the font, size, spacing, or text and background colours;
• hiding sidebars and headers, as browsers' reading modes do.

The Understanding document for 1.4.12 says people may set text spacing with "a user stylesheet, bookmarklet, extension, or application". A layout that cuts text off when spacing grows fails them, as button labels cut off when people widen text spacing shows.

A contrast theme goes further and replaces every colour. In Windows it is under Settings, Accessibility, Contrast themes, and browsers that follow it apply the theme's colours to the page, which CSS calls forced colors. Whatever the page shows only with a background fill, a box shadow or a fixed icon colour can disappear, as high contrast mode hiding focus, icons and selections explains.

People who don't use a mouse or a touch screen still have to reach every control. Each tool reaches it differently:

Tool	How it acts on a page	What the page needs
Speech recognition: Voice Control on iPhone and Mac, Voice Access on Android, Dragon on Windows	Say "Click" or "Tap" and the words on a control. With Voice Control and Voice Access, say "Show numbers" to label everything, or "Show grid" to point at a spot; Dragon's grid is "MouseGrid"	A name that starts with the visible words: buttons and links named differently from their text
Switches: Switch Control on iPhone, Switch Access on Android	Scanning highlights items one at a time; a press selects the highlighted one	Few steps to the content, and no action that needs a drag, a hover or a held gesture
Eye gaze: eye control in Windows, Eye Tracking on iPhone	An eye tracker moves the pointer, and holding the gaze on an item for a moment (dwelling) clicks it	Targets big enough to hit, and nothing that only appears on hover
Head pointers	Head movement moves the pointer, or presses keys on an alternative keyboard	The same as a mouse with one button, and every action from the keyboard
Keyboard only	Tab moves between controls, Enter and Space press them	A visible focus, a logical order, and a way past repeated blocks: skip links that are missing or can't be reached

Voice Control lists Japanese among its languages on Apple's availability page, and Voice Access lists it in Google's help; on a Japanese site, test speech in Japanese. Dragon's help describes clicking links and buttons by voice through its web extension, and its MouseGrid divides the screen into nine numbered areas that are redrawn smaller each time. W3C's resource adds that "not everyone using only the keyboard can see the screen", so the same keyboard path also has to be clear by ear.

Captions, transcripts and sign language serve people who can't hear the sound. W3C's resource describes captions as text of "the speech and non-speech audio information", synchronised with the audio, and notes that not all people who are deaf know sign language, and that there are many sign languages. The guides cover each: videos without captions, or with unchecked automatic ones, podcasts and silent videos with no transcript and videos without sign language, or with a tiny interpreter.

W3C's resource separates two things:

• Assistive technology: software and hardware such as screen readers, screen magnifiers, speech recognition and switches.
• Adaptive strategy: a technique people use, such as increasing text size, reducing mouse speed or turning on captions, with standard software, mainstream browsers, phones or assistive technology.

WCAG's definition of assistive technology notes that the distinction between mainstream user agents and assistive technology "is not absolute": many mainstream browsers and systems include features for people with disabilities. A page has to work with both kinds of tool, and with both ways of using them.

Skill varies as much as the tools. WebAIM's respondents rated their screen reader proficiency:

Proficiency	Respondents
Advanced	57.8%
Intermediate	38.8%
Beginner	3.4%

72.1% of advanced users go to the headings first on a long page, against 41.4% of beginners, and beginners are 2.7 times as likely to start with the Find feature. WebAIM notes that its sample was not controlled. Japan's National Diet Library said of its 2015 survey that many respondents were probably power users with high ICT skills. W3C's resource adds that tools may be unaffordable, not installed or not available in some languages, and that some people don't know a solution exists or how to set it up. So a page that works only for an expert who knows every shortcut doesn't work for many of the people using it.

EN 301 549's clause 4.2 describes the needs these tools serve, by what people can and can't do: usage without vision, with limited vision, without perception of colour, and eight more. Who is affected explains each group and lists the guides that reach it. How to test whether a technique works in the screen readers people use is on whether a technique works in the screen readers people use.