Test autocomplete accessibility with AI
Check the editable field, suggestion list and committed value together. Exercise keyboard selection, screen reader feedback and responses that arrive out of order.
Test autocomplete accessibility with AI when a search box offers suggestions but people using a keyboard cannot tell which value will be selected. The accessible-combobox-and-autocomplete skill identifies the intended pattern and checks its name, expanded state and active option. It also preserves ordinary editing behaviour in a field that accepts text.
A search for one prefix may finish after a newer search. If the old list replaces the new result, the visible options and announced active choice can diverge. Test that race alongside regular selection, dismissal and empty results, using a real browser and assistive technology for the behaviours they control.
What it does
- Check how the field announces its current state.
- Distinguish a highlighted option from a committed value.
- Test arrow navigation without breaking normal text editing.
- Compare pointer and keyboard selections.
- Check stale, loading and empty suggestion lists.
Install and use
- Install the skill and provide an authorised running combobox with representative test options. Use a keyboard, suitable browser and real screen reader for the relevant manual checks.
- Ask, “Review this product autocomplete. Change the query quickly, then select an item by keyboard and confirm what the field announces.”
- Repeat the interaction after any changes and record the environmental limits of the test.
FAQ
Source
Made by smukh / a11y-agent-skills contributors. Original skill: accessible-combobox-and-autocomplete, published on GitHub under the MIT licence.
View on GitHub