🎯Refining your Generated Prompt

Once your prompt has been generated, copy it to the prompt editor tab for further editing or customization.

Copy Prompt Over to the Prompt Tab

Step 1: Choose Copy generated prompt to copy the prompt to your clipboard.

Step 2: Return to the Prompt tab and paste the steps you copied.

In the prompt editor, feel free to customize the generated prompt to match your test case requirements. You can modify check assertions to verify alternative elements on the screen that the Prompt Assistant might not have initially selected. Additionally, you may want to amend the action command as needed.

Editing your Generated Prompt

Wait Command Insertion

When the app requires a few seconds to load a new screen, begin the prompt step with Wait until. This command instructs GPT Driver to pause until the app finishes loading, ensuring completion before proceeding to the next actions.

Replacing Coordinates

In specific cases, you might notice that the Prompt Assistant generates a prompt step related to Coordinates (i.e. x =, y =). This situation can occur when GPT Driver fails to identify a particular element or icon.

Since coordinates vary across different devices, it's advisable to substitute them with text. For example, if GPT Driver encounters difficulty recognizing a button and resorts to using coordinates, consider replacing them with the button's text label.

Although GPT Driver typically identifies most standard UI icons, it might occasionally struggle with elements or icons that are highly specific to your app. In such instances, you can create templates to address these challenges and replace Coordinates in your prompt. Explore more about Templates in our dedicated section.

Remember to save your test prompt frequently while editing. Exiting the test editor without saving will result in the loss of any recent changes.

Getting Tests to Reliability

Use the 10X feature, here is a quick video introduction on this feature

After creating the test, it is advisable to run it successfully at least once on the test editor to establish a baseline. For Android devices, you don't need to keep the browser tab in focus, allowing you to work on something else and then return to the GPT Driver tab once the test is completed, such as after 10 minutes. However, for iOS devices, we have observed that test execution in the editor will pause if the GPTD browser tab is not in focus, at least on some laptop and desktop devices.

The most efficient way of getting a test to reliability is to run them on the cloud. This way you can just set it and get back to it whenever. After getting the baseline, you can now run tests on the cloud.