Let the AI test your site
The AI doesn't just build — it checks its work like a visitor would. It writes a test suite for your site's important flows, runs it in a real browser against a throwaway copy (your live site and data are never touched), and reads the results — including screenshots — to catch what's broken.
What that means for you
When you ask for something that matters — a signup flow, a checkout, a booking form — the AI can prove it works end to end instead of assuming. If a test fails, it sees exactly where and usually fixes the problem in the same breath, before telling you it's done.
"Test the whole pre-order flow: pick a cake, fill the form, submit — and make sure the order actually shows up on my orders page."
"Something's wrong with signup — run the tests and fix whatever fails."
"Take a screenshot of the home page on a phone-sized screen and check nothing overlaps."
The Tests page
Your admin panel has a Tests page: the current test suite, past runs with their reports, failure details, and screenshots. There's a Run button to run everything on demand — a nice habit after a batch of changes. You never have to read the tests themselves; the reports are what matters.
Good to know
- Tests always run against a disposable copy with sample data, in a sealed sandbox. Real visitor data, real emails and real payments are never involved — payments and emails are stood in by simulators.
- Test runs have a generous daily budget; the AI spends them when they're worth spending, and you can always ask it to test more (or less).
- Screenshots from test runs are how the AI sees your site — which is also how it catches purely visual problems like unreadable text.