The short answer: prove the device is visible and watchable before you trust it with work
Before a new Android phone joins your daily task list, confirm two things in order: the phone farm software can actually find the device, and you can open a live view of its screen. Only after both are true should you consider assigning any real work. If either step fails, stop there and fix the connection rather than discovering the problem halfway through a task run.
This matters most right after you expand. A new batch of phones often arrives with a different model, a different Android version, or a different USB and network setup than the fleet you already trust. Two phones that look identical on the shelf can behave differently once connected, so the check is per device, not per shipment.
Why a new model or Android version is the risky moment
Your existing phones are known quantities. You have already seen them connect, you know their screens render as expected, and you know how they respond. A new model resets all of that. The same software that has been reliable for months may need a different setup step for a device it has never seen, and you will not know until you try.
The practical danger is not that the phone is broken. It is that a connection problem stays hidden until a task is already running, and then you are debugging under pressure while other phones wait. Checking availability first turns a mid-run surprise into a calm five-minute setup step.
Step 1: Confirm the device can be discovered
Open your phone farm software and look for the new phone in the device list. The goal here is simple: the software should register the device as present and available to work with. If it does not appear, do not move on to tasks. Resolve discovery first, because nothing downstream can be trusted if the software cannot see the phone.
When a device does appear, treat that as a starting signal, not a finish line. Being listed means the software found it. It does not yet prove the screen is usable or that a task will behave. That is what the next check is for.
Step 2: Open a live screen and look at what is actually there
Open the live screen for the new device and confirm you can see the current state of the phone. This is the check that separates a device that merely connects from one you can actually operate. Look at the screen the way you would look at a phone in your hand: is the display current, is it the screen you expect, and does it respond to what you do next?
In AnsClaw, this is where the AI side becomes concrete. Once the phone is explicitly selected and bound, you can describe a small, safe task in plain language and have the AI observe that one phone's screen and use its available tools to act on it, with you verifying the result. A sensible first action is something non-destructive and easy to judge, such as opening a settings screen or navigating to a visible menu item, then checking that the resulting screen matches what you asked for. If the screen state is unclear or a permission prompt appears, pause and confirm with a person before continuing rather than guessing.
Step 3: Decide whether this phone earns a place in the daily list
Only after discovery and a live, responsive screen should you decide whether the new phone joins routine work. If both checks pass, it has earned a trial. If either fails, keep it out of the task list until it is fixed. This is a decision point, not a formality, and it is cheaper to make it now than after a failed run.
One caution worth building into your habit: reconnecting is not the same as resuming. If a device loses its last connection, its window closes, and after it reconnects you should inspect what happened before rerunning anything. A reconnected phone does not mean an interrupted task picked up where it left off.
What this check does and does not tell you
A passed connection check tells you the software can find the phone and you can see its screen. It does not promise that every app on that phone will behave, and it is not a guarantee of stability across all models, Android versions, and network setups. Treat it as the minimum bar before real work, not a certificate of reliability.
This is also why the check is worth doing per device. A screen wall full of activity is not proof that tasks are succeeding; mature operations judge results per device rather than trusting a busy-looking display. The connection check is the first, cheapest version of that same habit: verify the specific device before you rely on it.
A realistic first session with a new phone
Put the steps together for one new phone. Add it, confirm it appears in the device list, open its live screen, and run one small, safe task through the AI on that single bound phone while you watch the screen and verify the outcome. If all of that works, you have a device you can trust with a trial task. If it does not, you have found the problem at the cheapest possible moment.
For operators comparing phone farm software, this order is a useful buying criterion too. Ask whether the tool lets you confirm discovery and a live screen before committing a device to work, and whether it keeps outcomes per device rather than showing only aggregate activity. A tool that makes the first check easy saves you from debugging blind later, and it fits how physical Android fleets actually get expanded: one unfamiliar phone at a time.