How to Build a Website with AI Without Coding — From a YouTube Video to a Live Site

Learning how to build a website with AI without coding is now less about finding the right tool and more about chaining several together. Julian Borge demonstrated the full sequence at Atlantis University's AI workshop: a 32-minute YouTube video he never watched became a fully functional website with working calculators, a dashboard and a learning center. He doesn't program. Ten months earlier, he didn't know how to do any of this. The entire build used free tiers.

The workflow, not the tool

The point Borge made repeatedly is the one worth leading with:

"The real skill with artificial intelligence is not knowing the different tools. It's creating the whole workflow."

Four steps, four tools, each doing one job:

Source material — a YouTube video, book, PDF, presentation or set of articles
NotebookLM — extract and organize the knowledge
ChatGPT — turn that knowledge into a professional build prompt
Lovable — generate the functional website

Interested in becoming a student?

Look for your passion in our available programs

Business School

Computer Science

Health School

Engineering School

A professional working on a laptop at Atlantis University

Step 1 — Pick your source

The demo used a video titled roughly how to build a 55-car rental fleet from scratch — 32 minutes long, and deliberately unwatched.

The workflow extracts and organizes the source material for you. Review the original context and verify the resulting claims before publishing.

Your source can be a supported video, book, slide deck, article or your own document. Google's standard Notebook plan currently allows up to 50 sources per notebook; check the current plan limits before building a larger collection.

Step 2 — Extract knowledge with NotebookLM

NotebookLM is Google's tool, which means YouTube integration is native — you can move directly from a video to a new notebook.

The process:

Copy the YouTube link.
In NotebookLM, create a new notebook and add the link as a source.
Ask targeted questions to pull out what you need.

The queries used in the demo were specific rather than general — the entrepreneur's journey, the business model, the biggest mistakes, the financial and operational lessons, website ideas. Specific questions produce usable material; "summarize this" produces something you'll have to rework.

Worth knowing: NotebookLM will also generate audio overviews, video overviews, slide decks, mind maps, reports, flashcards, quizzes, infographics and data tables from the same sources. For students facing a large final project, this alone justifies learning it.

Step 3 — Have ChatGPT write the prompt

Here's the single most useful piece of advice from the session, and it applies well beyond website building:

"Do not try to create professional prompts. Just ask the LLM for the prompt."

Many people stall because they believe they need to master prompt engineering. You don't. You need to describe what you want and let the model construct the instruction.

In practice, this happens in two passes:

First pass — before touching NotebookLM, ask ChatGPT what questions you should be asking about your source. It returns a set of extraction prompts. Take those to NotebookLM.

Second pass — bring NotebookLM's output back to ChatGPT and ask it to build a complete website specification prompt from that material.

What came back in the demo was substantial: an executive summary, product vision, business roadmap, a vehicle selection framework, a financial model, a complete sitemap, home page specifications and more. Borge's own assessment: "I wouldn't be able to create this by myself."

Step 4 — Build it in Lovable

Lovable builds functional web applications from natural language. Paste in the specification prompt, and it generates the site.

What "functional" meant in the demo — this wasn't a static page:

A homepage and business roadmap laying out the stages of starting the business
An explanation of the business model and vehicle categories
A startup cost calculator — working, showing what launching costs based on fleet size
A monthly profit calculator — working
A learning center and a fleet planning tool
A data dashboard

Every element operational, and the site publishable to a live URL.

Lovable can publish a project to its default subdomain. Custom-domain support and domain-registration costs depend on the plan and provider; check the current options before purchasing.

Free tier reality check. The demo used free tiers. Lovable uses credits and usage-based limits, so the number of iterations and the cost of running a live project depend on the current plan.

The advice on paying for tools

Borge was direct about spending, and it's the kind of guidance that's rare at an event with a commercial component:

"Before paying for any AI tool, test it and try it. Do not throw money into the garbage."

Everything in the demo was built free. He pays for ChatGPT and Claude because he uses them daily — but NotebookLM and Lovable were both free-tier throughout. Find out whether a tool is useful for your specific work before subscribing.

Ten months from zero

Worth stating plainly, because it changes how the rest of this reads.

Borge doesn't code. Ten months before this session, he couldn't do any of this. What changed was classes, and then — in his own words — a lot of testing and error over an extended period.

The workflow is genuinely accessible. It still requires practice.

Frequently asked questions

Can I really build a website without any coding knowledge?

Yes. Every tool in this workflow operates on natural language — you describe what you want.

Is Lovable free?

A free tier is available with limits. Check the current pricing page for included credits, hosting usage and custom-domain features before committing to a plan.

What are the alternatives to Lovable?

Base44 was mentioned as a comparable platform, and several tools now do similar work.

Can I connect a database or external services?

Yes. Lovable supports API connections, which is how you'd add authentication, payments or a CRM.

Can I use this for landing pages specifically?

Yes — it's arguably the best use case. If you want to capture leads and push them into a CRM, this workflow handles it well.

What's the difference between "Gemini notebook" and NotebookLM?

They're the same Google product, reachable from different entry points. Either works.

Conclusion

The reason how to build a website with AI without coding confuses people is that they go looking for one tool that does everything. There isn't one. There's a sequence — extract knowledge, generate a specification, build.

And that pattern isn't specific to websites. As Borge put it: if you work in lead generation, build a workflow. If you work in sales operations, build a workflow. The tools change constantly. The ability to chain them doesn't.

The Master in Artificial Intelligence at Atlantis University is built on exactly this principle — hands-on from the first session, adapted to your professional background.