Before you launch an app built with Lovable, Bolt, Cursor, or Base44, check that a stranger can do three things on the public URL: understand what the app does from the first screen, sign up without getting stuck, and reach the first useful result. Those three steps decide whether your launch traffic turns into users. This checklist covers each step, plus mobile, trust, and the quiet breakages AI-built apps tend to have.
Why is conversion the right lens for a launch?
A launch is a burst of attention: a Product Hunt day, a newsletter mention, a post that travels, an ad budget. That attention is expensive and rarely repeats. If the first screen confuses people or the signup fails, the burst is spent and the visitors do not come back.
Code that runs is necessary but not sufficient. The question for launch day is whether visitors become users. That is a conversion question, and testing is how you answer it before the traffic arrives.
The first screen
- The headline says what the app does and for whom, in plain words.
- The supporting line names the outcome a user gets.
- There is one obvious primary button, and its label says what happens next.
- The page has a meaningful browser tab title and a description for search and social previews.
- Header and footer links all load. Generated sites often keep placeholder links.
- The product name is consistent. Renamed products often keep the old name somewhere.
Signup and login
This is where AI-built apps most often break, because it touches email, redirects, and authentication settings that live outside the code.
- Signup works with a fresh email address, signed out, on the production URL.
- The verification email arrives, comes from a recognizable sender, and is not in spam.
- The verification link returns the user to the production app, not to a preview or localhost URL.
- Social login buttons, if present, complete and return to the right page.
- Password rules are stated before the user fails them.
- Error messages explain what to do, for example when an email is already registered.
- Log out, then log back in. It still works.
The first useful result
The first useful result is the moment a new user gets value: a first project, a first report, a first generated item.
- After signup, the user lands somewhere that tells them what to do first.
- Empty states explain the next action instead of showing a blank screen.
- The first core action works with realistic input, not just the demo data you used while building.
- Loading states show progress. A long silent wait looks like a crash.
- The result is clearly visible, and the user knows what to do with it.
Pricing and payment
- Pricing is visible before signup, or the signup page states what is free.
- Plan names, prices, and limits match everywhere they appear.
- The upgrade button leads to a working checkout in test mode.
- After a test upgrade, the account actually reflects the new plan.
Mobile
- The first screen, signup, and first action all work at a narrow phone width.
- Navigation menus open and close.
- Forms open the right keyboard and stay usable while typing.
- Nothing scrolls sideways.
Trust
- There is a privacy policy and terms page, and they load.
- There is a visible way to contact a human.
- The sender name on emails matches the product.
- If you collect sensitive data, the page explains why.
Quiet breakages to check specifically
- Environment variables or API keys that work in preview but are missing in production.
- Database rules that allow the builder's account but block new users.
- Redirect URLs still set to the preview domain.
- Rate limits or free-tier limits on the AI services your app calls, which can fail under launch traffic.
- Console errors on the public pages.
How do I run this checklist quickly?
- Deploy to the production URL you will share.
- Open a private window, signed out, on a phone width.
- Walk the first screen, signup, and first useful result, writing down every hesitation and error.
- Fix the issues closest to the money first.
- Walk it again before you share the link.
Persqo can run this walk for you with AI visitors. You paste your URL, edit the visitors so they match your audience, and each one tries to sign up and reach the first result in a real browser. You get screenshots, a session replay, and a fix prompt you can paste into Claude Code or Cursor. Start on the builders page, or run a free check. If you work from your editor, the developers page shows how to run it through MCP.
For the steps after signup, read from signup to activation and designing the first session.
The short version
- First screen: clear offer, one button.
- Signup: works on production, email arrives, links return to production.
- First result: reachable, with helpful empty states.
- Pricing: consistent, test checkout works.
- Mobile: the same path works on a phone.
- Trust: policies, contact, recognizable sender.
- Re-run after every fix and before you share the link.
Questions people ask
Why do AI-built apps need a conversion check before launch?
AI tools are very good at producing screens that look finished. They are less reliable at the paths between screens: signup emails, redirects, empty states, and mobile layouts. A launch sends paid or hard-won traffic into those paths, so checking them first protects your launch.
What is the single most important thing to test?
The path from your landing page to the first moment the product is useful to a new user. If a stranger cannot reach that moment, nothing else on the checklist matters yet.
Should I test in the builder's preview?
No. Test the deployed, public URL, signed out, in a private window. The preview often runs with your session, your data, and different settings than production.
How long does this checklist take?
Walking it by hand takes about an hour for a small app. An automated persona test can run the core paths in minutes and repeat them before every launch.