How I Built a TikTok Clone Without Writing a Single Line of Code Using AI

Hello everyone, I'm K.N, a developer who recently built a TikTok‑style website entirely with AI tools, writing no code myself.

AI has sparked a debate: some see it as a savior that lets developers sip tea while tools do the work, others fear it will replace developers. I take a balanced view—AI is a powerful tool, but its usefulness depends on how we use it.

Since first encountering GPT in late 2022, I’ve used it as a partner for repetitive tasks, code generation, writing, and even complex module creation.

Below I share how I used GPT, Cursor, and simple screenshots to quickly launch a short‑video site without writing a single line of code.

Preview URL: https://tiktok-clone-ashy.vercel.app/

1. Prompt Improvement

Accurate prompts are key to success. Although Cursor’s Composer Agent can generate projects from natural language, we still need to refine prompts so Cursor understands our requirements and produces high‑quality code.

Key points for better prompts:

Identity setting

Clear requirements

Task breakdown

Context usage

Iterative refinement

1.1 Identity Setting

First, I define an identity for GPT, e.g., an AI Prompt Engineer, to give it a background and capabilities relevant to prompt optimization.

<code>// prompt
I want to do [], please create a fictional [] identity background including education, research area, teaching experience, academic achievements, and industry experience.
[Extension]

// example
I want to write a prompt that helps Cursor understand, please create an AI Prompt Engineer identity background with education, research area, teaching experience, academic achievements, and industry experience.
</code>

GPT then generates a detailed identity description, which I prepend to my conversation.

<code>I need you to adopt the following identity and assist me with AI prompt optimization:
... // generated identity information
</code>

1.2 Clear Requirements

The requirement is simple: I upload a screenshot, and GPT analyzes the UI layout to generate prompts that help Cursor build the page. The description must be as detailed as possible to avoid misunderstandings.

For brevity only a portion of the reference prompts is shown; see the original article for full details.

<code>// prompt
During the process, always follow these steps:
1) I will upload an image (UI design or website screenshot) and tell you whether the target is a web app or mobile app.
2) Analyze the UI, layout, styles, and overall structure.
3) Provide step 1: initial page prompt, and step 2: page structure analysis.
4) Wait for my command before proceeding to step 2.
5) Include technology stack, import conventions, component organization, etc.
</code>

After testing with an Apple homepage screenshot, the AI successfully analyzed the layout and generated useful prompts. I then tried a more complex site (TikTok) to demonstrate the workflow.

2. Rapid Project Construction with Cursor

After installing Cursor (see its website for details), I opened the Composer with ⌘/Ctrl + I and entered a prompt to initialize an empty project.

It’s important to enable Agent mode so that terminal commands can be executed.

After initialization, I sent the TikTok screenshot to GPT, received refined prompts, copied them into Cursor’s Composer, and let Cursor generate the UI. Within seconds, a TikTok‑style front‑end appeared.

The result was a static site with interactive navigation, though videos were placeholders. By swapping images with video assets, the site became fully functional.

2.2 Generating Additional Routes and Pages

I fed the step‑2 prompts to Cursor, which automatically created secondary pages such as profile, settings, login, and live streams.

The generated pages demonstrated how AI can handle complex front‑end structures with minimal human intervention.

3. Deployment

Sticking to the “no‑code” principle, I asked Cursor to upload the project to GitHub and deploy it on Vercel.

<code>Please help me upload to GitHub and deploy to Vercel hosting service.</code>

The AI guided me through the necessary commands, repository creation, and login steps.

After following the prompts, the site was live.

Conclusion

AI has evolved from an abstract concept to a practical assistant that can accelerate development and handle repetitive work. However, it remains a tool; its value depends on the developer’s judgment and critical thinking. Developers must continue to learn, evaluate AI outputs, and avoid over‑reliance.

The core competitive advantage of developers stays in judgment and thinking ability , not in the tools they use.