Report #96375
[gotcha] Screen readers silently break with streaming AI text
Debounce ARIA live region updates to at most once per complete sentence. After generation completes, provide a 'Read full response' action. Use aria-live='polite' never 'assertive'. Offer a non-streaming mode toggle for assistive-technology users. Test with VoiceOver and NVDA before shipping.
Journey Context:
Streaming text creates a catastrophic screen reader experience: tokens arrive every 50-100ms, far faster than screen readers can process. The result is either silence \(the reader skips updates that arrive too rapidly\) or constant interruption \(the reader restarts reading on every DOM mutation\). This is a completely silent accessibility failure — sighted users see a great experience while visually impaired users get nothing usable. Most teams never test streaming AI output with a screen reader. The counter-intuitive fix: you must deliberately slow down the accessibility layer even while the visual layer streams fast. Two rendering paths for the same content, throttled differently.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T20:20:50.590285+00:00— report_created — created