Mobile UX for AI Copilots: Small Screens, Big Expectations

Desktop copilots get keyboards and wide context windows. Mobile copilots get interrupted attention, one hand, and impatience. Designing AI for phones means ruthlessly prioritizing input, output, and interruption.

Input: voice, chips, and keyboard

Offer voice where privacy allows, plus tappable prompt chips for the top three tasks. Keep the composer sticky and avoid hiding send behind nested menus. Autocomplete should never fight the user's cursor.

Output: stream with structure

Stream text for perceived speed, but chunk into paragraphs and bullets as content arrives. Long walls of tokens feel broken on mobile. Use collapsible sections for citations and sources.

Mobile users treat AI like messaging — they expect instant feedback and clear 'still working' states.

Handle interruption

Save thread state, support background resume, and never lose partial input on app switch. Network drops should offer retry without wiping context. Haptics and subtle animation confirm sends without slowing power users.

Connect to native UI

Deep-link AI suggestions into existing screens — maps, calendars, product detail. The copilot is a launcher; your proven UI still closes the loop.