Testing your app with AI

WeWeb AI can try the app it builds. It takes over the editor in preview mode, opens your pages, types in forms, clicks buttons, and watches what happens, just like a user would. When something is wrong, it tells you what it found and fixes it.

This saves you the back and forth of testing a form yourself, describing what went wrong, and waiting for a fix. It also catches problems you might not spot right away, such as a submission that fails without any message, or text that is the same color as its background.

Starting a test

WeWeb AI never starts a test on its own. There are three ways to ask for one:

• Click Run AI test on the Done card at the end of a build. The button shows up when the AI changed something in the Interface tab that has not been tested yet.
• Ask for it in the chat, for example "Can you test the signup form?"
• Report a problem, for example "The signup button does nothing". The AI tries to reproduce the problem before fixing it, instead of asking you to test it yourself.

Be specific about the page, the flow, and what you expected. The AI tests what you ask for, not the whole app, so a precise request gets you a faster and more useful test.

Letting the AI take over

Before it touches anything, the AI asks for the editor with a Let the AI test your app? card in the chat. The card says what it wants to test and why.

• Click Let it test to start the test.
• Click Not now to refuse. The AI tells you it could not test, without pretending anything passed.

If you do not answer within 15 minutes, the request expires and the AI tells you it could not test. You can ask again whenever you are back.

THE TEST USES YOUR REAL APP

The AI acts on your app for real: forms get submitted and records may be created in your tables. Read Test data to know how to find what it created.

GET NOTIFIED WHEN THE AI IS READY

It is easy to switch tabs while the AI builds and miss the request. In the chat settings, set Sound when generation is done to On (under Notifications): you hear a sound when the AI asks to test, and if you are on another browser tab, a browser notification tells you to click Let it test.

What happens during the test

Once you accept, the editor switches to the Interface tab in preview mode, wherever you were before (even in Data & API). You can tell the AI is in control:

• A green frame surrounds the page.
• A Live testing badge replaces the preview controls in the top bar.
• A green cursor moves to each element the AI acts on.
• The chat shows the test plan as a checklist, with each step ticked off as the AI goes.

While the test runs, the AI can:

• Open the pages of your app
• Type in inputs like a user, so workflows that run when a value changes (live validation, search suggestions) run as well
• Click buttons and links, which runs the workflows attached to them
• Scroll to load content further down the page
• Switch to the mobile or tablet view to check the layout
• Read error messages, variable values, and the logs of the workflows it triggered

It also spots layout problems from the page itself: elements that overflow or overlap, truncated text, and text you cannot read because it has too little contrast with its background.

To keep the test stable, you cannot leave preview mode, open the Data & API or Settings tabs, or close the chat panel until it ends.

Stopping a test

Click the chat's Stop button at any time. The editor goes straight back to the tab and mode you were on before the test, and the AI stops.

After the test

When it is done, the AI gives the editor back and reports:

• What it tested
• The problems it found, most serious first
• What works
• The test data it created
• What it could not check, and why

Then it moves on to fixes:

• If you reported a problem, the AI fixes it directly.
• Every other problem appears in a question where you choose what to fix. The fixes the AI recommends are already checked. Click None of these to leave everything as it is.

After the fixes, the AI does not rerun the whole test. The Run AI test button comes back on the Done card, so you can test the changes again when you want.

Test data

Since the AI submits real forms, it can leave accounts or records behind. To make them easy to find, it uses names and emails that start with qa_test_ followed by a timestamp, and lists everything it created in its report.

The AI does not delete this data on its own. When the data is clearly disposable, such as an account created to reproduce a bug, it may add a "Delete the test data" option to the question of fixes. Otherwise, remove it yourself from your tables when you no longer need it.

TEST ON DATA YOU CAN LOSE

If your app is connected to data that matters, such as real customers or payments, keep that in mind before clicking Let it test. The AI follows the same flows as your users, so a checkout flow or an email workflow runs for real.

Where live testing is available

Live testing works with WeWeb AI in the editor, in every workspace. It needs the editor on the other end, so it is not available to external agents connected over WeWeb MCP.

Learn more about the AI modes →