BLOG
BlogGuides

Testing an app you built with AI: a guide for founders

You built it with Claude Code, Cursor, Lovable or Bolt and it looks done. How to check it works for a stranger and keeps working after every change.

Done is not the same as working#

Building with AI is fast enough that testing becomes the slow part, so it gets skipped. The cost shows up later: agencies that rescue AI-built apps describe the same failures again and again, from the blank screen in production to the fix that breaks something else.

This guide collects what we have written for people building this way. Each post is about one problem, says why it happens, and gives a plain-language test you can run against your live address. Where we have a real run to show, the post links to it.

1. Check what the agent called done#

2. The flows that break for strangers#

3. Keep it working#

4. Let the agent test its own work#

5. Know the limits#

The proof behind the posts#

A post that says Testsy can do something should show it doing it. The runs linked from these posts are real runs on our practice site, with their results, replays and videos, open to anyone until 18 February 2027. The one in the picture at the top is a planted bug: an export that said it had started and downloaded an error page instead of a CSV. The tester opened the file and failed the check, in 40 seconds (the run).

Start with one test#

Paste your app's address into New test, describe sign-up and one thing your app saves, and run it. The free trial has 10 runs and needs no card.

Filed underWebAI-built apps