Launch checklist for an AI-built app

The week before

Domain

• The app runs on your real domain over HTTPS.
• The www and bare domain both work, and one redirects to the other.
• Auth redirect URLs, OAuth apps and webhook endpoints use the real domain, not the preview one.

Email

• Your sending domain has SPF, DKIM and DMARC records, so password resets and receipts don't land in spam.
• You've sent a real sign-up and password reset to a Gmail and an Outlook address and checked they arrive in the inbox.
• Replies to your emails reach someone.

Payments

• Stripe, or your provider, is in live mode with live webhooks registered.
• You've bought your own product with a real card, checked that access was granted, then refunded it.
• Your receipts and statement descriptor show a name customers will recognize.

First-run experience

• You've signed up as a new user on a phone and a clean browser and gone all the way through the core feature.
• Empty screens explain what to do next.
• Test accounts, placeholder text, debug banners and “coming soon” links are gone.

The basics people check

• A privacy policy and terms that describe what the app actually collects and does. Get advice if you're unsure what applies to you.
• A support email that someone reads, linked from inside the app.
• A page title, description, share image and favicon.
• No leftover noindex tag from a staging setup, and a sitemap if you want search traffic.

Measurement

• Analytics records sign-ups and payments, and you've seen the events arrive from production.
• Error tracking is on, and you've seen a test error appear.

Launch day

• Launch early in a day you can watch, not late on a Friday.
• Deploy before you announce, and run through sign-up and payment one more time.
• Keep the error tracker and logs open.
• Know how to roll back, and don't ship new features that day.
• Check your providers' limits: email sending quotas, API rate limits, and free-tier database or function limits.

The week after

• Read every new error. Early errors tend to point at real flows users are stuck in.
• Look at where people drop off between sign-up and first use.
• Talk to the first users who stay, and the first who leave.
• Ship small fixes often rather than one large release.
• Check your bills: AI model usage, SMS, email and hosting.

If something breaks and more prompting isn't fixing it, see when AI coding gets stuck.