Guidepup
Open-source driver to automate tests on real screen readers (NVDA on Windows, VoiceOver on macOS) from Node.js. Lets you assert what the end user actually hears, not just what axe-core measures. 2026 reference for voice/screen reader E2E a11y tests. Single multi-screen-reader API: if NVDA or VoiceOver have a keyboard command, Guidepup supports it.
Strengths
- Real screen readers (NVDA, VoiceOver), not a mock or emulator
- Single multi-screen-reader API, learn once, write for both
- Covers UX a11y bugs invisible to machine checks (reading order, applied ARIA)
Limitations
- Heavy CI, requires machines with audio access (Windows for NVDA, macOS for VoiceOver)
- No JAWS support (40% desktop market), gap for heavy enterprise products
- Fragile tests if synthesizer voice changes or OS updates, maintenance required
Best for
- Products with strong a11y obligation (gov, banking, EAA June 2025) where screen reader UX matters
- Mature a11y teams that already have axe-core in CI and want to go further