Mobile AI Testing Should Focus on Input Capabilities, Not Device Type

A new framework argues that testing AI interfaces on mobile devices should be based on active input capabilities and state transitions rather than broad device categories like phone or tablet. Modern devices blur these lines — foldables can change screen size mid-task, tablets may combine touch with keyboard and stylus, and voice input can lose microphone permission unexpectedly. The protocol recommends logging detailed context including device model, OS build, input hardware, network conditions, posture, and permission state to make test results reproducible. Key metrics to capture include time to first streamed content, recovery after resize or resume, duplicate submissions, and lost draft characters. The author, who contributes to the open-source MonkeyCode project, clarifies this is a test protocol rather than a device benchmark, and no performance claims are made about any specific product.