An AI tool built me a website and gave me files. Now what?

This page may contain affiliate links from which Blur Studio earns commissions.

This is a genuinely new version of an old problem. AI coding tools are good at producing working website code fast, and bad at explaining that "working on my screen" and "live on the internet" are two different things. The files exist. They just aren't anywhere the public can reach them.

First, figure out what kind of site the AI actually built

This matters more than which AI tool you used. There are two very different cases:

• Plain static files: HTML, CSS and JavaScript that run entirely in the visitor's browser, with no login system, no database, and no server-side logic. A portfolio, a simple business site, or a basic landing page usually falls here.
• A fuller application: anything with user accounts, a database, saved data between visits, or server-side code that needs to run somewhere other than the visitor's browser. Many "AI app builder" tools produce this kind of project by default, even when you only asked for a simple site.

If you're not sure which one you have, check whether the AI mentioned a database, an API, environment variables, or a backend at any point. If none of those came up, you very likely have the simpler case.

For plain static files, this is the easy case

Drag the folder of files onto Netlify Drop and the site is live within seconds, with no account needed for that first deploy. This is the fastest path from "AI gave me files" to "it's actually online," and it costs nothing. Create a free Netlify account afterward if you want to keep the site and connect your own domain.

One thing worth knowing if you're iterating with the AI tool and redeploying often: Netlify's Deploy Previews now generate automatically not just from pull requests in a connected Git repository, but from AI agent runs that change files directly, each getting its own preview URL in the pattern agent-{RUN_ID}--yoursite.netlify.app. If your AI tool supports connecting through Netlify's agent integration, you can see a live preview of each AI-made change before deciding to publish it, rather than redeploying blind every time you ask for a tweak.

For a fuller AI-built app, this gets more involved

If the AI tool generated something with a database, user accounts, or server-side logic, a plain file drop genuinely won't work, since there's nothing to run that server-side code or store that data. Your real options at that point:

• Vercel is built around exactly this kind of project (particularly apps built with frameworks like Next.js, which many AI coding tools default to), with a comparable Git-connected deploy flow to Netlify's own.
• Hire a freelancer to deploy it makes sense if the AI-generated code has genuine bugs or missing configuration that needs a developer's eye before it can run anywhere reliably, which is common with AI-generated backend code specifically.
• A proper custom build is worth considering if the AI output is close but not quite right, rather than paying repeatedly to patch something that was never built with your actual requirements in mind.

Netlify itself can still host the front-end half of a fuller app, connected via Git rather than a plain file drop, with its functions and database features covering some backend needs; whether that setup fits depends on exactly what the AI built, which is specific enough that it's worth checking case by case rather than assuming one answer.

Step by step: the fast path for static files

1. Confirm there's no database or server code mentioned anywhere in what the AI gave you. If there is, skip to the fuller-app section above instead.
2. Open index.html locally in a browser first. Confirm it actually works before publishing it; AI-generated code sometimes references files or links that don't exist.
3. Drag the whole folder onto Netlify Drop. Not individual files; the whole folder.
4. Check every page the AI generated, not just the homepage. AI tools sometimes generate convincing-looking pages that reference other pages that were never actually created.
5. Create a free account and claim the site if it works. This keeps it live permanently and lets you connect a real domain.

When to skip this entirely

If the AI project needs a database or server code, a plain file drop will not be enough, no matter how carefully you follow the steps above; you'll just get a broken site that looks right locally and fails online. This is the single most common frustration with AI-built sites, so check for it before spending time on a deploy attempt that can't work.

A worked example: a chatbot-built landing page that looked done

Say a solo founder asks a general-purpose AI chat tool to "build me a landing page for my app, with a signup form that saves emails to a list." The AI returns HTML and CSS that render perfectly in the chat tool's own preview, including a form that appears to work. What the founder doesn't immediately realize: the "saves emails to a list" part was never actually implemented, the form has no destination, it was described in a code comment rather than built as working code. The site looks completely finished right up until someone tries to sign up after it's live, and nothing happens on the business side even though the visitor sees a success message.

This is a specific, recurring failure mode worth checking for directly rather than assuming "it looked done in the AI tool" means "it's actually done." Before publishing anything with a form, a login, or any action that's supposed to save or send data somewhere, test that action and confirm you can see the result arrive, an email in your inbox, a row in a spreadsheet, a message in a Slack channel, not just a success message on the page. A form that shows "Thanks, you're signed up!" without actually recording the signup anywhere is arguably worse than no form at all, since it gives both the visitor and the business false confidence that something happened.

Common mistakes with AI-built sites

• Trusting a feature because the AI described it confidently. AI tools frequently describe a feature as implemented in the same breath as actually building it, and the two aren't always the same thing. Confidence in the explanation isn't evidence the code behind it works.
• Not asking the AI directly whether anything needs a server or database. A simple, direct question, "does this site need a database or backend server to work, or is it pure HTML and JavaScript that runs in the browser," often gets you a clearer answer than inferring it from the code yourself.
• Publishing before testing every interactive element. Static text displays fine almost by default. Forms, buttons that are supposed to do something, and anything claiming to save data are exactly the parts that silently fail and need direct testing before anyone else sees the site.

How I'd approach this

When someone brings me an AI-built project that's "almost there," the first thing I check is exactly this: is it actually a simple static site that just needs hosting, or does it need real backend work the AI glossed over. If it's the simple case, I tell people to just deploy it themselves, since paying for help with a drag-and-drop deploy is money better spent elsewhere. If the AI output has real gaps, that's where I come in: finishing a half-built AI project properly, with code you actually own, rather than patching around missing pieces indefinitely. Get in touch if you're not sure which situation you're in.