From ChatGPT or Codex to your own website: publishing AI code in production

Published on 2026-10-27 by frfrand

How to transform an app written with ChatGPT/Codex into a real site: structure, builds, GitHub, API, VPS, SSL and testing.

Can an app written by ChatGPT be published?

Yes, as long as the result is an executable project with the necessary dependencies. But it is not the same to receive a single HTML file as it is to receive a system with users, payments, and a database. An AI can create both; production requirements change depending on what the software does. You do not need a complex server if all you are delivering is a static landing page.

How to recognize what type of application the AI gave you

If you received HTML/CSS/JS without a backend, a static server may suffice. React/Vite requires compiling and serving files. Next.js with API routes requires server-side execution. A Python or Node API requires a persistent runtime and configurations. Identify scripts, language version, folder structure, and external connections. Everything that persists users or imports must have a data model and backups.

Example: budget calculator

ChatGPT generates a calculator that adds hours and materials: it works in your browser, but when you reload, budgets are lost. If the goal is to save jobs for clients, define accounts, tables, permissions, and API. The full flow is to create a budget, save it, retrieve it from another session, and export it. Only then do you choose the database and publish the backend behind HTTPS. A screenshot does not prove that the budget was saved.

How to deploy it to the server

1. Place the files in a repository with installation instructions. 2. Run the build from scratch with fixed dependency versions. 3. Replace mock data and simulated functions that affect the business. 4. Add login, persistence, and error tests. 5. Choose the appropriate file server, runtime, or container. 6. Configure domain, HTTPS, error logging, and data restoration.

Does ChatGPT replace infrastructure?

No: it generates ideas, explanations, and code, but the web needs a place to run and services to keep it available. A serious deployment also requires a person responsible for responding to failures.

Frequently asked questions

### Do I have to buy a domain immediately? No: test on a temporary URL before making the final cutover. ### Can I copy the code without review? It is not prudent if there is real data or payments; verify license and security. ### Do Codex and ChatGPT always generate the same framework? No: it depends on the prompt, the repository, and the tools used. ### How much does migration cost? It is determined after identifying backend, data, functions, and required support level.

Sources

OWASP Top 10 · GitHub Actions · Vite Build.

Moving from AI code to ArduMaker's migration service

The goal of the future service at ardumaker.com/lovable is to study AI-generated code, adapt it to a working deployment, move services if agreed upon, and publish it on controlled infrastructure. The page does not exist yet and we are not announcing an available automatic migration tool. At ArduMaker you can check the current VPS and services infrastructure. Each project requires diagnosis: a static landing page, an SSR with a database, and a SaaS that charges subscriptions do not need the same process.

Editorial note: October 2026. Test first in staging, document licenses, and keep verified copies before modifying production.

What the photograph shows

Programmer reviewing generated changes. Illustrative photograph generated to accompany this guide on migrating applications created with AI; it is not a screenshot of the platform nor does it certify a specific installation.

Continue with these practical ArduMaker guides

- Taking Cursor-created code to production: from editor to an app that doesn't crash - Deploying an app made with Claude Code: GitHub, servers, domains and backups - React + Vite created with AI: publishing a SPA with Nginx, SSL and domain

If you are evaluating your own application, compare the real backend, database, authorization and deployment requirements before choosing the server.