Gemini and Claude are now handling routine phone tasks — reordering food, adjusting smart home devices, cleaning Google Drive, but only within tight guardrails. Android Police’s eight-month test shows neither AI is ready to run your phone unsupervised. They excel at rigid, repeatable chores. They fail at nuance: tone in email, cross-document reasoning, open-ended browsing.

And they cost extra: Gemini’s screen automation caps at five free requests per day. Full agentic automations require Google Home Premium or the Advanced plan. Claude, meanwhile, handles calendar edits and Drive cleanup without subscription walls, but can’t trigger actions on-device like Gemini can. Neither replaces human judgment.

Both demand verification. Your mileage will vary. But if you hate reordering the same DoorDash meal every Tuesday, one of them just might do it for you, then wait for your thumbprint before charging.