WebAIM published its 11th Screen Reader User Survey on 29 September 2026. It ran in July and August 2026 and had 1,780 responses. One number stands out: 91.8% of respondents use AI in at least one way.
How people use it
- 60.1% use AI to generate image descriptions or alt text.
- 49.7% use it for complex images such as charts.
- 40.6% use it for guidance on online tasks.
- 37.1% use it to summarise pages.
Apple's May 2026 accessibility announcements point the same way: VoiceOver can now describe images with Apple Intelligence, and new features summarise and simplify text.
The tempting conclusion, and why it is wrong
It is tempting to read this as "users have AI now, so alt text and structure matter less". The survey says the opposite.
- People reach for AI because sites leave images undescribed. A model guessing what a chart means is a workaround, and it can be wrong in exactly the places that matter: a price, a dosage, a trend line.
- 67.8% still navigate by headings. No assistant fixes a page whose headings are missing or out of order; it can only describe the mess.
- Only 32.2% think the web became more accessible in the past year, and 22.9% think it got worse.
Your alt text is the description the author intended. An AI description is a stranger's guess. Give people the first, and they can still use the second when they want to.
Which screen readers to test with
Respondents' primary screen readers were JAWS (55.0%), NVDA (32.9%) and VoiceOver (6.5%) for this survey's audience. NVDA is free, so there is no cost barrier to testing your key pages with it.
What to do
- Write alt text for informative images, and say what the image is for in context. For charts, put the key numbers in text near the chart.
- Use real headings in a logical order (
h1toh2toh3), because that is how most people move through a page. - Test one key journey with a screen reader each release. Ten minutes with NVDA finds problems no checklist will.
Sources: WebAIM: Screen Reader User Survey #11 results and Apple Newsroom, May 2026.