ChatGPT Sites: Wait, I Could Build That Here?

I just found out about Sites in ChatGPT.

Yes, the guy with a website about building things with AI missed a feature for building things with AI. Please allow me a moment to adjust my very professional neon sunglasses.

My first thought was about this website. I could use something like this to give a little project its own home, then link to it from Vibe Code with Matt so people can actually try the thing.

Then I saw the collaboration features. Now I’m thinking about all the small tools a team could use but nobody ever has time to build.

This is my introduction to Sites and what I want to try with it. The hands-on project report comes after the hands-on project. A bold editorial policy, I know.

What is ChatGPT Sites?

Sites lets ChatGPT build, host, and update websites, web apps, and games from a prompt or a compatible project. Hosting means the result has somewhere to live online. It is in public beta for Plus, Pro, Business, Enterprise, and Edu, with plan-specific limits. OpenAI’s Sites documentation covers current availability.

The appeal for me is the distance between “I have an idea” and “here, click this.”

What can you build with Sites?

Start with something you can explain in one sentence and check yourself. “Help me figure out how much paint I need” is a better first assignment than “manage my entire home renovation.” The kitchen can wait.

Here are some starter ideas for your first ChatGPT Site:

A calculator for something you keep looking up

A paint calculator could take a room’s wall measurements, subtract doors and windows, and estimate how much paint you need. A recipe scaler could turn ingredients for four servings into enough for seven. A running pace calculator could convert a distance and finish time into minutes per mile.

Tell ChatGPT the inputs, units, and calculation you want. Ask it to show its assumptions beside the result. For paint, that might mean coverage per can and number of coats.

Your own shopping list app

Start with adding an item, changing its quantity, and checking it off. Group items by aisle if that matches how you shop. Make the buttons big enough to use on a phone while steering a cart with your other hand.

Decide what “save my list” means before building. Keeping it in one browser is a smaller project than syncing it between your phone and a partner’s phone. Put that requirement in the prompt so you can test the right behavior.

Leave barcode scanning, receipt uploads, and an AI meal-planning empire for later. First, make sure you can remember the eggs.

A simple game with one good idea

Try a memory matching game, a short trivia quiz, or a reaction timer. For a first version, ask for one screen, clear rules, a score, and a restart button.

Give it a personal theme: matching your pets’ faces, trivia for a family gathering, or a quiz about a hobby you know far too much about. Knowing the subject makes it easier to spot when the AI invents an answer.

Test what happens when someone taps twice, gets every answer wrong, or restarts halfway through. Players are very good at finding the button sequence you didn’t think of.

A small tool for an everyday routine

A packing checklist, a study timer, a garden watering log, or a random dinner picker can be useful without becoming a whole business.

Pick the one thing that makes it yours. A packing checklist might change based on trip length. A study timer might include the break schedule you actually follow. A dinner picker might exclude anything that requires washing three pans.

If you’re stuck choosing, use the guide to picking your first vibe coding project, or browse the project gallery for inspiration. Aim for a problem you recognize and a result you can judge.

The team collaboration part got my attention

Where available, owners can grant editing access to active members of the same workspace through Share, changing Can view to Can edit. Editors can save changes and publish updates after the owner’s first publication. Owners retain access management. Editors can read the live database. External invitations are for viewers, not editors, and are rolling out separately. OpenAI explains these permissions here.

For a team, consider an onboarding checklist, a meeting agenda builder, or a content launch hub with a brief, owner, and next step for each piece of work.

Choose one workflow and ask the people doing it what gets in their way. Build around that answer. After a week, an editor could add a missing field or simplify a layout based on actual use.

The goal is to make the job easier without accidentally appointing someone full-time caretaker of a prettier spreadsheet.

Agree on who handles publishing before starting, and use made-up data in the first version. Learn whether the workflow helps before moving real work into it.

How to build your first Site

Open Sites in ChatGPT, or mention @Sites in a prompt. Save a version for review before deploying; deployment produces a live site. See the getting-started guidance.

Give ChatGPT a specific job, a short feature list, and a clear description of how you’ll use the result. Here is a shopping list prompt you can adapt:

@Sites Build a simple shopping list website for me to use on my phone.

Let me add an item and quantity, edit or delete it, and check it off.
Group items into Produce, Dairy, Pantry, and Other.
Keep unchecked items above checked items within each group.

Save the list in this browser so it survives a refresh or closing
the tab. Explain that this version will not sync across devices
and that clearing browser data can remove the list.

Use large tap targets, readable text, and labeled controls that
also work with a keyboard. Do not rely on color alone for checked items.

No accounts, sharing, or meal planning for this first version.
Start with an empty list and a helpful message explaining how to add items.
Save a version for me to review before deploying.

Check the ordinary stuff before adding more features

For this shopping list, add eggs, change the quantity, check them off, and delete them. Try a long item name and an empty entry. Refresh the page with a few items still on the list, then close and reopen it in the same browser.

Open it on your phone too. Can you tap the checkbox without accidentally deleting dinner?

Ask someone to try it without coaching. Watch where they hesitate. That tells you more than asking whether they like the colors.

Ask for one specific improvement at a time

Describe the actual problem: “The delete button is too close to the checkbox on my phone. Separate them and ask me to confirm before deleting an item.” Small, specific changes are easier to judge than “make it pop.”

I have asked AI to make things pop. Some of them are still popping.

When to keep the scope small

Keep your first project easy to check and rebuild. A recipe calculator or memory game fits that description. Something handling customer records or charging money calls for a different level of planning.

Sites’ current restrictions include financial transactions and payment-card data. Check the current limits and restrictions before planning a business app.

For any tool that saves work, test closing it, reopening it, and checking that the right person can still see the right information. A nice-looking Save button is only half the assignment.

As the project grows, write down what you need to own and maintain: the content, source code, domain, and any stored records. Those requirements should drive the tool choice. My OpenAI Codex page and Tools by Job hub are useful starting points when figuring out the rest of a project’s workflow.

Pick something you’ll actually use

Scale a recipe. Finish a round of trivia. Get through the grocery store with everything you came for. Those are perfectly respectable goals for a first app.

If you’ve been sitting on an idea, write down the smallest thing someone could actually use. Start there. You can add the sprawling feature list after a real person asks for feature number two.

Meanwhile, I will be investigating what else is apparently sitting right in front of me in ChatGPT.

The sunglasses may be part of the problem.