What's built into the interface
The application is designed and tested against WCAG 2.2 AA. In practice that means keyboard operability throughout, visible focus indicators, sufficient colour contrast, adequate target sizes, and content that reflows without horizontal scrolling when zoomed.
- Full keyboard operation without a mouse
- Screen reader support across the interface
- Contrast and target sizes meeting AA thresholds
- Reflow at high zoom without horizontal scrolling
Lemon says: If a student reports something that isn't keyboard-reachable or a screen reader can't announce, please report it — that's a bug, not a limitation.
Features that help more students than you'd expect
Several conversation controls were built for accessibility and turn out to help a much broader range of students — anyone anxious, anyone in a noisy room, anyone at a lower proficiency level.
- AI Text — the AI's turn shown as text, which helps with hearing difficulties, processing time, and simply missing a word
- Pause and Resume — time to think without the session running on, unless the activity limits or turns off pauses
- Push-to-talk — control over exactly when the microphone is listening
- Speaking pace — a slower AI is easier to process for many students, not only beginners
Lemon says: Setting pace to 0.85× helps far more students than the ones with a formal accommodation.
Activity settings you can use as accommodations
Several accommodations can be delivered by building a second copy of an activity with different settings, rather than by exempting a student from the work. This is usually a better outcome for everyone.
- Extra processing time — a copy with slower pace and a longer maximum time, or, where pauses are limited, a Pause allowance for that student
- Reduced pressure — a copy with a lower level and stronger scaffolding
- Anxiety — completion-only assignment, no rubric, and explicit permission to retry
- Fatigue — a copy with a shorter minimum time so a session is achievable in one sitting
Lemon says: Copy the activity, change the settings, assign the copy. The student does the same task at a workable difficulty.
Students who cannot do spoken work
Some students have accommodations that make spoken assessment inappropriate — speech disabilities, severe anxiety, or a documented exemption. The platform is a speaking tool, so the accommodation has to come from your course design rather than from a setting.
- Substitute an equivalent written or asynchronous task for the same objectives
- Assess the same objectives in a one-to-one meeting instead
- Where possible, keep the activity available ungraded so the student can still use it if they want to
Lemon says: Keeping it available but ungraded is often the kindest option. Some students want to try without the stakes.
Speech recognition and speech differences
This is the accessibility issue most specific to speaking practice, and it deserves stating plainly. Speech recognition is measurably less accurate for speakers with accents, speech differences, and lower proficiency — which means the students most likely to be misread are those already most disadvantaged.
Lemon says: When a transcript looks wrong for a particular student, listen to the audio before you act on it.
Working with disability services
When your institution's disability services office asks about the tool, these are the points they generally need. Send them the accessibility statement, and the conformance report (available from ChitterChatter on request), alongside your own course-level plan.
- The application targets WCAG 2.2 AA
- It runs in a browser with no separate software install
- Speaking activities require a microphone, and text alternatives are available for the AI's turns
- Instructors can vary pace, level, session length, and scaffolding via activity copies assigned to a section, and extend pause time per student
- Substitution remains available where spoken assessment is inappropriate
Lemon says: Answering this proactively in week one is much easier than answering it in week nine with a deadline pending.

