Anyone who needs a screen reader needs it for accessing the whole device so building a screen reader into a browser is kind of pointless. The exception being Go

Anyone who needs a screen reader needs it for accessing the whole device so building a screen reader into a browser is kind of pointless. The exception being Google's ChromeVox but that's because on Chromebooks the browser effectively is the OS. NVDA and Windows's Narrator are already probably good enough but JAWS is better. VoiceOver is built into macOS and iOS and is beyond good enough. Browsers can do a better job of exposing semantics through accessibility APIs to assistive technology [0]. Browsers could also intelligently make up for the failings of web sites; e.g. when a site uses a div with a click event instead of a button element, present it as a button through the accessibility API based on heuristics. Browser rendering engines already do a lot to visually compensate for errors in HTML, they could do the same for some semantic errors. [0] http://www.html5accessibility.com http://www.html5accessibility.com