Build a website with AI from 0 to 1: go live for free
A flow I actually ran end to end: talk the idea through with AI, have it write the docs, generate the code, run end-to-end tests and deploy — using only WorkBuddy plus DeepSeek, with money spent only on a domain (which is optional). Nine steps, and you have a live site that search engines and AI can find.
A lot of people think building a website means knowing how to code, renting a server and wrestling with toolchains. The bar today is absurdly low: you mainly need to be able to describe what you want clearly, and let AI do the rest. The flow below is one I actually ran end to end — from a one-line idea to a live site that can be searched and cited by AI — using only WorkBuddy plus DeepSeek, with money spent only on a domain (which you can skip too).
1. Talk your idea through with AI
The first step is not writing code — it is talking the idea through with AI. You do not need to say everything at once: go back and forth like a chat. Start with roughly what you want to build; the AI asks questions you had not thought of, you answer, it asks more — until you think "yes, this is it".
2. Have AI write a spec and an architecture doc
Once the idea is clear, do not rush to ask for code. Have AI turn the requirements into a spec (what features, roughly what the pages look like) and then an architecture doc (which stack, how the folders are laid out, where the data lives).
This step is easy to undervalue: it turns the fuzzy idea in your head into an explicit agreement on paper. From then on you and the AI work against the same document, so nothing drifts as the conversation goes on.
If the look matters to you, have AI generate a reference mockup first (using GPT image generation), agree on the style, and only then let it write code. Redrawing a picture is far cheaper than rewriting a pile of code.
3. Create a Gitee repository
The code needs a home. This step is one action: create a repository on Gitee. Pick Gitee because it is fast to reach from here, free, and private repos are unlimited — ideal for a personal project just starting out.
git init
git add .
git commit -m "init: spec + architecture docs"
git remote add origin https://gitee.com/<your-account>/<repo-name>.git
git push -u origin mainYou do not type these by hand — hand them to AI and let it initialize and push for you.
4. Commit the docs to the repo so nothing is lost
The spec and architecture docs are not throwaway files you close and forget — put them in the repo alongside the code.
Three wins: it cannot be lost, it keeps a version history (you can see what changed), and a different AI can pick up from the same docs later. This step costs almost nothing, yet losing the docs means losing all the thinking you already did.
5. Have AI generate the project code
With the docs ready, let AI write the code against the spec and architecture. It creates the folders, the pages, the styles and the data files by itself — you just review as it goes.
6. Have AI run end-to-end tests
Code that is written is not code that works. Have AI run end-to-end tests: open the page in a real browser, click through it, measure the key elements and capture a few screenshots for you.
End-to-end tests catch whether the whole thing works, not whether one line is correct — they stop problems like "this button does nothing" before you ever notice them.
7. Deploy to Vercel — ride the free tier
Deploy on Vercel: connect the Gitee repo and every push deploys automatically, with HTTPS certificates issued and renewed for you. A personal project will never exhaust the free tier — it costs nothing.
After deploying, Vercel hands you a xxx.vercel.app address that already works over HTTPS — effectively a free domain. If you want your own name (say a .com), buy one from a registrar like NameSilo, then let AI wire the DNS to Vercel; certificates stay automatic.
8. Connect Google Search and keep tuning SEO
Once live, go to Google Search Console and submit your sitemap so Google knows you exist and which pages you have.
SEO is not a one-off. Have AI fill in title, description, canonical and structured data on every page, then watch Search Console: whichever page is not indexed or whichever keyword gets no clicks is where you improve.
9. Have AI optimize for GEO
More and more people ask AI directly instead of typing into a search box. GEO (generative engine optimization) is about making AI cite your site.
- Write content as one question plus one direct answer, no adjectives — AI quotes the sentence that answers the question.
- Use structured data (JSON-LD) to state plainly what the site is and what it contains, so AI does not have to guess.
- Provide a plain-text summary for AI (such as llms.txt) so it can read your whole site the easy way.
The heart of this flow is not any technology but two small things: say it clearly, and keep a record of every step. Do those two and AI does the rest — between an empty repository and a freely hosted site that search engines and AI can find, there is little more than nine rounds of conversation.