Screen reader

In practice

It does not read your CSS or your HTML the way you wrote them. What it works from is the accessibility tree the browser builds out of them, with some products also reaching into the document itself for things the tree does not carry. That is why what gets announced can differ from what you wrote, and why a page that validates cleanly can still say nothing useful.

Users navigate by structure rather than reading top to bottom, and the structure they reach for is headings. Asked what they do first on a long page, seven in ten move through the headings, one in seven uses the browser's find, and fewer than one in twenty-five starts with landmarks. Landmarks are still worth having, since about a third of users say they use them often. Headings are where the effort pays first, and skipped levels are the defect to look for.

Braille is not a footnote. Around two in five of the survey's respondents with disabilities read some of the page on a braille display, where speech rate does nothing, audio cues do nothing, and structure carries the entire experience.

Most of the testing done in this industry is desktop-shaped and much of the use is not. Nine in ten of those respondents also use a screen reader on a phone, and only four in ten mostly use a desktop, with half saying they use both about equally. A growing majority say they would rather use an app than a website for a common task. If your testing stops at a laptop, it stops before the audience does.

Which one you test in changes what you find, and it changes by region. In that survey JAWS led as the primary reader in North America and Australia, and NVDA led in Europe, Asia, Africa and the Middle East, in one case by three to one. Testing only in JAWS is not therefore a test for North America alone, since plenty of people there use something else. What it does mean is that a global audience is a reason to widen the matrix.

Why it matters

Alt text, accessible names, heading structure, landmarks, live regions and reading order are all answers to the same question. What does this sound like? Two findings from the users themselves are worth carrying into any conversation about budget. Asked whether better websites or better assistive technology would improve accessibility more, around 86% said better websites, and that share has risen every time the question has been asked since 2009. And two thirds of them never or rarely contact a site owner about a barrier, which means nobody has complained is not evidence of anything.

Where this shows up on the site

• How our blind testers work
• JAWS
• NVDA
• VoiceOver

Related terms

• NVDANVDA is a free, open-source screen reader for Windows, made by NV Access, and more people reach for it than for anything else.
• JAWSJAWS is the long-established commercial screen reader for Windows, made by Freedom Scientific, and the one named most often when people are asked which screen reader they use first.
• VoiceOverVoiceOver is Apple's screen reader, built into every Mac, iPhone and iPad, and into the Apple Watch, Apple TV and Vision Pro as well.
• Braille displayA braille display is a hardware device with cells whose pins rise and fall to form braille characters, refreshing as the reader moves through a page.
• Accessibility treeThe accessibility tree is a second structure the browser builds alongside the DOM, holding one object for every part of your page that has to be exposed to assistive technology.
• Assistive technologyAssistive technology is hardware or software that gives a person with a disability more than a mainstream browser offers on its own.
• Browse modeBrowse mode is the state a Windows screen reader sits in while somebody reads your page, and forms mode is the state it switches into when they type.