I Built This Website With ChatGPT Sites. My Honest Review

What ChatGPT Sites actually gives you

OpenAI describes Sites as a way to create, host, refine, and share websites, web apps, and even games from inside ChatGPT.

For me, the easiest way to think about it is this:

It combines a coding agent, a website preview, and managed hosting inside one conversation.

I can explain what I want in plain English, review what it builds, send screenshots when something looks wrong, and keep asking for changes until it feels right.

I do not have to set up a separate hosting account or spend time connecting several different tools before I can see the website running.

That is where most of the value comes from.

It is much more flexible than I expected

The biggest surprise has been how much control I still have.

I am not limited to choosing from a few templates or dragging blocks around a page. I can ask for custom layouts, interactions, forms, animations, blog systems, newsletter signups, and tracking.

For this website, I wanted more visual movement, so I added Three.js. That is not something I expected to be able to do easily inside an AI website builder.

I have also been able to work with tools such as Cloudflare Turnstile, PostHog, and Google Analytics. Sites supports hosted environment variables and secrets, which makes these types of integrations much more practical.

Sites also provides its own basic analytics. It automatically tracks unique visitors and page views, so you may not even need another analytics tool if you only want a simple view of your traffic.

For deeper tracking, I still like having the option to add something like PostHog or Google Analytics.

This does not mean Sites can support absolutely everything. OpenAI’s documentation clearly says that some frameworks, databases, background services, and hosting patterns are not supported.

Still, it feels much closer to working with a real code project than working with a closed drag-and-drop website builder.

Hosting is probably the easiest part

Once the website is ready, Sites handles the hosting and gives you a live URL.

You can also connect an apex domain or subdomain that you already own. You update the DNS records through your domain provider, and the website continues to be managed through ChatGPT.

This works, but it could be made much easier. A better experience would be a simple “Connect domain” button that automatically adds the required DNS records for providers like GoDaddy, Hostinger, or Cloudflare, similar to how Resend handles domain setup. Instead of manually copying records, you would just authenticate your provider, approve the changes, and everything would be configured automatically.

My own website is an example of that.

This removes a surprising amount of work. I do not need to think about building the website in one place, moving it to another service, configuring a deployment, and then connecting the domain separately.

I can focus on what I want the website to say and do.

For a personal website, blog, portfolio, landing page, or quick experiment, that ease of use is a major advantage.

The production workflow still needs work

The biggest weakness for me is the lack of a traditional environment setup.

To be fair, Sites does provide private previews and saved versions. You can save a version, review it, and deploy it when you are ready.

But every actual deployment URL is a production deployment.

That is not the same as having separate development, staging, and production environments. In a normal software workflow, I may want my team to test a change in staging, approve it, and then promote that exact version to production.

Sites gives me a useful review step, but it does not yet feel like a complete staging and release process.

For a personal website, I can live with that.

For a larger product or a website where a mistake could affect revenue, I would want stronger controls.

I want a clearer GitHub workflow

My second issue is code ownership and visibility.

OpenAI supports working from a local source project. Its documentation also says that saved versions can be associated with Git commits, and that local projects can be edited and tested through Codex tools.

That is useful, but the experience is less clear when the website starts entirely inside ChatGPT.

I would like a simple option that says, “Connect this website to my GitHub repository.”

From there, I should be able to review every change, inspect the code, make a manual SEO update, run my own tests, or move the website to another hosting provider if I ever need to.

Right now, I can ask ChatGPT to make SEO changes, update metadata, improve performance, or fix technical issues. But that is different from having the repository directly available and knowing that I can take over manually at any time.

The capability may exist through a more local, developer-focused workflow. I just think the web-first Sites experience should make that path much more obvious.

It still makes mistakes

Sites is powered by AI, and AI sometimes gets things wrong.

A change that sounds simple can affect another part of the page. A design may look correct on desktop but break on mobile. A form may need another round of testing. Occasionally, I need to send a screenshot and explain the problem more clearly.

The speed of Sites can actually make this more dangerous because it becomes tempting to publish changes quickly.

You still need to check the copy, links, forms, tracking, mobile layout, and interactive behaviour before publishing. OpenAI recommends the same review process in its official Sites guide.

Sites makes building faster. It does not remove the need for judgment.

It is also still in public beta

ChatGPT Sites is currently in public beta.

Availability depends on your plan, region, and workspace settings. Usage limits can also change, and certain technical setups are not supported.

That does not stop me from using it, but I would keep it in mind before building something business-critical on the platform.

Who I think ChatGPT Sites is best for

Sites makes a lot of sense for:

• Personal websites and blogs
• Portfolios
• Landing pages
• Internal tools and dashboards
• Small experiments
• Founders who want to launch quickly
• People who have ideas but do not want to manage the complete technical setup

It may be less suitable if you need a formal release process, complex infrastructure, strict compliance requirements, or a large engineering team working through pull requests and multiple environments.

My current verdict

My experience with ChatGPT Sites has been very good.

Its biggest strength is not that it can magically build the perfect website from one prompt. One prompt is rarely enough.

The real strength is how much distance it removes between having an idea and putting that idea online.

I can think of a change, describe it, review the result, and publish it without turning every update into a separate development project.

Would I use it today for every complex software product?

Probably not.

Would I use it for a personal website, blog, landing page, portfolio, or fast experiment?

Yes.

In fact, I already am.

This website is the proof, and so is Ciqora, the product I’m building. Its waitlist site is another example of what I’ve been able to create with it.

I will keep documenting what works, what breaks, and what I learn while using AI in real projects. If that sounds useful, you can join my newsletter and follow along.