When an AI system is added to a service, it can change what information people see, how they interpret it and what they believe they should do next. Cognitive accessibility belongs in those design decisions from the start.
Can people understand the response?
Review the length, structure and language of AI output in the context of the task. Give people a way to ask for clarification, revisit earlier information and find the next action without having to reconstruct a long conversation.
Is uncertainty communicated clearly?
An answer that sounds certain can invite a decision even when the system lacks enough information. Explain important limits in plain language and make it clear when a person should verify a claim or seek human support.
Who retains control?
Test whether people can correct assumptions, review consequences and undo or pause an action where appropriate. The aim is a system that supports understanding and judgement, especially in journeys with real consequences.