How a 14-Year-Old Built a Website With Claude Code

Last updated: July 2026

Kids building websites with AI is no longer a hypothetical, and one of my students just proved it. A 14-year-old whose parents bought her a Claude subscription used the Claude Code command-line tool to build her own portfolio website, put it on GitHub, and deploy it live to the web. She did the specifying and the deploying herself, with my supervision at each step.

This is not a story about magic. It is a story about what a ready teenager can actually do when you give her a real tool and stay in the room. Below I will walk through exactly what she built, what she genuinely learned, what I supervised, and how you could set up the same path for your own child if they are ready for it.

Key Takeaways

• A supervised 14-year-old built and deployed a real portfolio website using Claude Code, GitHub, and Vercel, with me guiding the steps.
• The real learning was not "the AI wrote code." It was breaking work into steps, reading and approving each change, and understanding version control and deployment.
• Readiness matters far more than age: this worked because she already had coding foundations and the patience to read what the AI proposed.
• We are holding off on buying her a domain until the site is fully functional, and I will not publish her live URL while consent is still pending.
• A second student is now on the same path, which tells me this is a repeatable setup for the right kid, not a one-off.

Table of Contents

• What She Actually Built
• What She Genuinely Learned
• What I Supervised, and Why
• Is Your Child Ready to Try This?
• How to Set Up the Same Path
• Related Articles
• Frequently Asked Questions
• The Bottom Line

What She Actually Built

She built a personal portfolio website, the kind a young creative or future developer puts their projects on. A homepage, an about section, a place to show her work, clean navigation, and a layout that looks good on a phone.

She used Claude Code, Anthropic's command-line coding agent, to do it. Her parents had already bought her a Claude subscription, so the tool was sitting there waiting. What turned it from a chat toy into a real project was structure.

We set her up with a GitHub account so her code lived in a proper repository, and we deployed the site to the web using Vercel, which connects to GitHub and publishes the site automatically. Once the site is fully functional, the plan is to buy her a domain so it has a real address.

Her father was genuinely happy, and I understand why. There is something different about a kid pointing a browser at their own live website versus showing you a screenshot of an app on a learning platform.

What She Genuinely Learned

This is the part parents should care about, because it is easy to look at a finished website and assume the AI did everything.

She learned to break a big goal into small, shippable steps. "Build me a website" is not a thing you can do in one move. "Add a navigation bar with three links" is. She got better at this every week, and that skill transfers to almost any kind of work.

She learned the basics of version control and what GitHub is for. Every change she kept became a saved version she could go back to. That is a concept most adults never learn, and she now has a working feel for it.

She learned to deploy something to the public internet, which means she understands, at a kid's level, the difference between code on her laptop and a site the world can load. And most importantly, she learned to read a proposed change and decide whether to approve it. Claude Code shows you what it wants to change before it changes anything. She had to look, understand roughly what it was doing, and say yes or no.

That last habit, reading and approving rather than blindly accepting, is the entire difference between a kid who learns and a kid who copy-pastes.

What I Supervised, and Why

I was not building this for her, and I was also not sitting back and letting the AI run unattended. The honest answer is somewhere in the middle, and that middle is where the learning happens.

Here is roughly how the supervision broke down across the work.

Task	Who did it	My role
Deciding what to build next	The student	Asked her to be specific before she started
Writing the prompts to Claude Code	The student	Helped her sharpen vague requests into clear ones
Reading the proposed code changes	The student, with me	Slowed her down so she actually read, not just clicked yes
GitHub setup and first deploy	Together	Did the account and config plumbing with her watching
Debugging when something broke	The student leads	Pointed her at the right file, let her drive the fix

The pattern that matters: she made the decisions, and I made sure she understood them. When a change looked wrong, I did not fix it. I asked her what she thought it was doing. That question, asked at the right moment, is something no AI tool can do, and it is most of what a good tutor is for.

A second student is now on the same path, also building a portfolio site with Claude Code. The fact that it transferred to a different kid tells me this is a setup, not a fluke.

Is Your Child Ready to Try This?

Readiness is the whole game here, and it has almost nothing to do with whether your child is exactly 14.

The student who pulled this off already had coding foundations. She could read a small piece of code and have a rough idea of what it did. She had the patience to look at a proposed change instead of mashing accept. And she wanted to build something real, which gave her a reason to push through the boring parts.

If your child cannot yet write and explain a small program on their own, a tool like Claude Code will skip past the layer they actually need to learn. They will produce a website they cannot read or change, which feels like progress and is not. I wrote about this readiness threshold in detail in How to Know Your Kid Is Ready for Python, and the broader picture of what AI literacy means for kids is in Can Kids Learn AI? What Parents Need to Know.

One of my students' parents, Matt, once described my role as molding a kid's coding skills rather than just handing them answers. That is exactly the distinction that decides whether a project like this teaches anything.

How to Set Up the Same Path

If your child has real foundations and you want to try this, here is the order I would do it in.

First, confirm readiness honestly. Can they write a small program and explain it? If not, build that first with structured lessons before any AI agent enters the picture.

Second, set up the accounts properly. You own the Claude subscription, not the child. Put it on a shared computer where you can see the screen. Create a GitHub account and a Vercel account together so the publishing pipeline is ready.

Third, pick one specific, small thing to build. "A portfolio site with a homepage, an about page, and a projects section" is buildable. "Make me a website" is not.

Fourth, make a rule that every proposed change gets read before it is approved. This single rule is what converts the tool from an answer machine into a teacher. Sit with them for the first several sessions until reading-before-approving becomes a habit.

A quick note on safety and consent: do not publish your child's live site URL or their domain publicly until you have agreed, as a family, that it is appropriate. I am keeping this student's site address private for exactly that reason.

Related Articles

• I Gave an 11-Year-Old a Claude Code Subscription, the younger-kid version of this same supervised-AI approach.
• Can Kids Learn AI? What Parents Need to Know, the pillar guide on what AI literacy actually means for a child.
• How to Know Your Kid Is Ready for Python, the readiness check that decides whether a tool like Claude Code will help or hurt.

---

Frequently Asked Questions

Can a 14-year-old really build and deploy a website with Claude Code? Yes, with supervision and existing coding foundations. My student specified what she wanted, read and approved each change, and deployed the result to the web using GitHub and Vercel. The AI handled the heavy lifting of writing code, but she made the decisions, which is where the learning lives.

Did the AI do all the work for her? No, and this is the key distinction. Claude Code shows proposed changes before applying them, so she had to read and approve each one. My job was to slow her down so she actually understood what she was approving rather than clicking yes on autopilot.

What did she learn that will last? Breaking a big goal into small steps, the basics of version control with GitHub, what it means to deploy something to the public internet, and the habit of reading a change before accepting it. Those skills transfer well beyond this one website.

Is my younger child ready for this? Probably not yet if they cannot write and explain a small program on their own. Readiness matters far more than age. Build foundations first with structured lessons, then revisit an AI coding tool once they can read and reason about code.

Will you share the student's website address? No. Consent is still pending, and as a rule I never publish a student's live URL or domain. We are also waiting until the site is fully functional before buying a domain for it.

What does this cost to set up? The student's family already had a Claude subscription, and both GitHub and Vercel have free tiers that cover a personal portfolio site. The real cost is the human supervision, your time or a tutor's, which is the part that makes it educational.

---

The Bottom Line

A ready 14-year-old can build and deploy a real website with Claude Code, but the result is only educational when an adult keeps her reading and approving instead of blindly accepting. The tool did not teach her. The combination of a real tool, a real project, and a human asking the right questions did.

Wondering whether your child has the foundations to try a project like this, or whether they need to build them first? Book a free Discovery Call and we will assess where they actually are and map the right next step.