🔤AI Instructions

In essence, AI instructions guide GPT Driver's interactions with the device via natural language.

AI instructions vary in complexity. The simpler ones handle basic tasks, like tapping buttons, scrolling screens, or entering text. Additional instructions handle more advanced tasks such as Network Requests.

Instruction	Expected Action	Example
Tap on	Taps on specified element, text, or icon.	Tap on text
Type	Types out specified text	Type text
Scroll (Up, Down)	Scrolls up or down on the center of the screen in chosen direction	Scroll down, one screen at a time, until you see text
Swipe (Left, Right)	Swipes on the center of the screen in a chosen direction	Swipe left on the screen
Slide (Left, Right, Up, Down)	Slides specified element to the left, right, up or down	For swiping to any direction, the best prompt that works consistently uses this format: Slide direction, by XX% on the element. Where XX% is an estimate of how long the swipe should be in relation to the screen's width/ height.
Check	Checks screen to see specific element or text. Will fail the test if specified element or text is not visible.	Check that you are one a screen that displays text
If - Otherwise	First checks screen to see specific element or text. If visible, next action will be carried out. If not, you can direct GPT Driver to carry on with the test.	If text is displayed on the screen, then tap on text otherwise continue on to the next step.
RemoveText	Removes existent text (only within an input field)	RemoveText: text
Remember	Remembers a specified element, text, icon or previous action taken.	Remember the text visible on the screen … Type the remembered text into the input field
Task Complete	Directs GPT Driver to succeed the test under specified condition(s)	Task Complete: when text is visible on the screen
Error Detected	Directs GPT Driver to fail the test under specified condition(s)	Error Detected: when text is not visible on the screen
Execute: network request		Execute: network request <GET/DELETE>;url=;json selector=
Open Deeplink	Opens specified Deeplink on device	Open Deeplink: link
Double Tap	Taps twice on specified element, text, or icon.	Double tap on text
Tap in sequence	Fast taps x times in sequence on element, text or icon	Tap in sequence 7 times on text
Verify Email		Verify email: email=;link_pattern=<link_pattern>
Long Press	Holds down on specified element, text, or icon e.g. to bring up Copy/Paste option	Long Press on text
Continue with next step	Continues with next step in the Test Prompt	Continue with next step
Press Enter	Presses the Enter key	Press Enter
Go to Device Home Screen	Directs the app to the device home screen	Go to Device Home Screen
Shake the device	"Shakes" the device (only available for iOS)
withVision	Enables enhanced visual assertions, such as verifying colors and visual attributes. And we now also support performing actions based on visual cues.	withVision: Verify if the current screen background is color red. withVision: Tap on any red Live button on the screen