OpenAI launches Accessibility Mode in ChatGPT with simplified language and semantic cues

AI · 4 min read

Accessibility Mode provides an alternate response format that favors short, structured paragraphs; explicit headings; and optional inline descriptions of charts and images. It also labels interactive suggestions and buttons in a way that improves readback by screen readers. The mode can be set as a user preference or applied at the conversation level, and it supports easy toggling for developers embedding ChatGPT in products.

Technically, the mode uses an internal response template layer combined with a post-processing pass that injects ARIA‑friendly metadata and alternate text cues. OpenAI says it worked with accessibility researchers and user groups, including blind and low-vision testers, to iterate on phrasing, verbosity, and the handling of complex content like code snippets and tables.

Accessibility Mode highlights a broader shift in generative AI toward offering outputs tailored for user needs rather than forcing a single 'default' style. Advocates welcomed OpenAI’s move but urged transparent documentation and interoperability — for example, explicit signals in the API so downstream apps and assistive tech can detect and consume accessible output programmatically.