You made your project with an AI tool and it looks good in the preview. Now you want to publish it and it turns out “publish” is not a button: the tool gave you the code, but not the place where it has to run.
1. Find out what kind of project it is
Before choosing hosting, look at what it is made of:
- Frontend only (static): HTML, CSS and JavaScript. Easy to publish on services like Cloudflare Pages.
- Frontend with its own backend: it needs a server, a database and environment variables.
- Frontend with external services: it uses a database or authentication such as Supabase or Firebase. Their access rules need to be reviewed.
2. Separate testing from production
What works on your computer uses test settings. In production you need:
- Your own environment variables (API keys, database connection), written outside the code.
- A production database separate from the test one.
- A domain with HTTPS and the DNS records set up correctly.
3. Make publishing repeatable
If you publish by uploading files by hand, sooner or later something breaks. The best setup is for every change to go through the repository and publish itself: it is built, tested and always comes out the same.
4. Backups and alerts from day one
Before launching, find out where the database backups are and who finds out if the site goes down. A downtime alert and a daily backup solve most problems.
5. When to ask for help
If the project handles customer data or payments, or needs to grow, it is worth having a developer review the deployment and the security before launch. Fixing it afterwards costs more.
If you want me to review yours, see how I take projects made with AI to production.