QA testing a vibe coded app before launch
QA testing is someone using your app the way real people will, including the ways nobody planned for, and writing down every place it breaks. AI-built apps pass the demo and fail on the edges, which is exactly what QA is for.
QA testing an AI-built app costs $1,500 to $8,000 CAD for most small products, and it finds the bugs your first users would otherwise find for you: the refund that runs twice, the form that accepts a negative quantity, the page that breaks when a name has an apostrophe, the flow that dead-ends when someone presses back. AI builders optimize for the path you described in the prompt. QA walks every other path.
What QA finds that a demo does not
| Type | Example |
|---|---|
| State bugs | An order marked paid before the payment clears, or stuck in "processing" forever after a failed card |
| Double actions | Clicking submit twice creates two bookings, two charges or two accounts |
| Boundary values | Zero, negative, very large and empty inputs accepted where they make no sense |
| Broken navigation | Back button, refresh or a deep link lands on an error or an empty screen |
| Data edge cases | Accented names, long text, emoji, time zones and daylight saving, currency rounding |
| Role mix-ups | A feature built for admins appears for everyone, or a user loses access after an update |
| Integrations | Email, payment and calendar steps that fail silently and leave the record half-written |
| Mobile and browsers | Layouts that work on the builder's preview and break on a real phone or on Safari |
QA and security testing are different jobs
QA asks whether the app does what it should for a person using it in good faith. A security test asks what a person acting in bad faith can make it do. They overlap at the edges, since a logic bug in checkout is both, but a QA pass will not try to read another tenant's data and a security test will not check that your date picker handles leap years. Most founders need both before a launch with real money or real personal information. They are cheaper bought together. See the security test page for the other half.
How QA testing runs
- List the flows that matter: sign-up, the core action your product exists for, payment, cancellation, account deletion.
- The tester builds cases for each: the happy path, then the edges, then the combinations.
- Testing across the browsers and devices your users actually have.
- A report with each bug, the steps to reproduce it, what should have happened and a suggested fix. See what the report includes.
- You fix, usually by pasting the bug into your AI tool, and the tester confirms the fix.
Can you launch without it
You can, and plenty of apps do. What you are choosing is who finds the bugs. For a free tool with no accounts, early users finding them is fine. For anything that takes money, holds personal information or sells to businesses, the first bug a customer sees costs more than the test. The trade-off is set out on launching without QA.
What it costs
| Scope | Typical range |
|---|---|
| Core flows, one role, desktop and one mobile browser | $1,500 to $3,500 |
| Several roles, payments, email and one integration | $3,000 to $6,000 |
| Multi-tenant app, several integrations, full device matrix | $5,000 to $8,000 |
Bundled with a security test, the combined engagement usually lands at $5,000 to $12,000 CAD for a small app. TrazTech sells the two together as vibe-coding QA and security review, listed from $2,000 CAD.
What you can do yourself first
Run the launch checklist before paying anyone. It catches the obvious gaps, and a tester's time is then spent on the ones you could not find. The launch readiness check scores where you stand in two minutes.
Get QA before your users do it for you
List your core flows and who uses them. You get a scope and a price range back.
Get matchedCommon questions
Can the AI write my tests instead?
It can write automated tests, and you should keep them. They test what it thinks the app should do, which is the same understanding that produced the bug. A human tester brings a different model of the product.
How long does QA take?
Two to five working days of testing for a small app, plus a day for the report. Retesting fixes takes a day or less.
Do I need QA if my app is free?
Less urgently. If it has accounts and stores anything personal, the security side still matters even when nobody pays.