Digital interfaces are often explained visually: click the icon in the top-right corner, or drag the blue panel. For someone using a screen reader, even finding help can become another barrier. AskEase, a Microsoft Research prototype, addresses this problem with step-by-step guidance designed around non-visual computer use.

What AskEase does

The assistant combines several sources of context to understand the user's goal and current situation. Instead of offering a generic tutorial, it creates a sequence of actions for the precise point where the person is stuck.

Instructions are written for screen-reader navigation: they refer to labels, navigation order and useful commands without relying on the visual position of elements. The interaction is designed to interrupt the main task as little as possible.

  • On-demand help based on the current context
  • Steps written for non-visual navigation
  • Less dependence on immediate human assistance
  • Guidance powered by language models

What the screen-reader user study found

The evaluation involved twelve screen-reader users. Researchers report that AskEase improved task success and reduced perceived workload, including effort and frustration. The sample is small, but the results suggest value in AI that does more than describe the screen: it explains how to act within an interface.

The project also addresses independence. Tutorials and videos are frequently designed for sighted users, while an experienced helper is not always available at the moment a problem occurs.

A research prototype, not an available product

AskEase is a research project and should not be presented as a Microsoft product ready for daily use. Wider deployment would require testing across more applications, languages, experience levels and assistive configurations.

One principle must remain clear: AI may help people work around an inaccessible interface, but it must not excuse software without labels, semantic structure or keyboard controls. Contextual assistance works best on top of products that are already designed accessibly.