Apple brings AI to its accessibility tools — richer descriptions, live subtitles and eye-controlled wheelchairs

On May 19, 2026, timed to lead into Global Accessibility Awareness Day, Apple previewed a suite of accessibility features that weave its on-device artificial intelligence, Apple Intelligence, into tools that millions of people already rely on every day. The theme running through the announcement is quiet and practical: use AI to make existing features do more, without sending users’ most personal information off their devices.

For blind and low-vision users, VoiceOver gains an Image Explorer that produces detailed descriptions of photos, scanned bills and personal records, and a Live Recognition mode that lets a person point the camera at the world and ask questions about what it sees, with natural follow-ups. Magnifier adds a high-contrast interface controlled by voice, so someone can say “zoom in” or “turn on flashlight.” Voice Control introduces a “say what you see” approach, letting users navigate with natural phrases such as “tap the purple folder” instead of memorizing exact labels.

“The theme running through the announcement is quiet and practical: use AI to make existing features do more, without sending users’ most personal information off their devices.”

Two features stand out for how directly they widen access. On-device generated subtitles automatically transcribe spoken audio in uncaptioned videos across iPhone, iPad, Mac, Apple TV and Vision Pro — a boon for deaf and hard-of-hearing users, and all processed locally rather than on Apple’s servers. And for people with limited mobility, Vision Pro’s eye-tracking can now steer compatible power wheelchairs over Bluetooth, launching with Tolt and LUCI systems in the United States — a first for the headset.

“Now, with Apple Intelligence, we are bringing powerful new capabilities into our accessibility features while maintaining our foundational commitment to privacy by design,” said Apple CEO Tim Cook. The emphasis on on-device processing matters here: accessibility tools handle deeply personal data — medical documents, private conversations, the contents of a person’s home seen through a camera — and keeping that computation local is itself an accessibility and dignity feature.

Apple says the new capabilities will arrive later in 2026, with generated subtitles beginning in U.S. and Canadian English. As previews, they still have to prove themselves in daily use across the enormous diversity of how disabled people actually live and work. But the direction is heartening: some of the most meaningful uses of AI are not flashy, but the steady expansion of what a blind, deaf or mobility-impaired person can do independently, with tools already in their pocket.