I Built 3 Websites with AI-Assisted Development: Here's What I Learned

Boyan — Mon, 09 Nov 2026 10:13:05 +0000

I've always been interested in creating useful websites. But there's one thing that makes my journey a little different: I don't come from a traditional programming background.

Instead of starting with years of coding experience, I started with ideas, curiosity, a willingness to learn, and AI-assisted development tools.

Over time, I've been building and managing three independent websites: Zearches, MeWebLink, and ChipJourney.

Each project has taught me something different about website development, automation, security, SEO, and the challenges of running online platforms.

What surprised me most wasn't how quickly AI could help generate code. It was how much work happens after that code is written.

Here's what I've learned along the way.

1. Zearches: Building a Free Website Discovery Directory

My main project is Zearches, a free website submission and discovery directory.

The idea came from a simple frustration: getting a website listed in online directories can sometimes involve fees, reciprocal backlinks, lengthy approval processes, or complicated registration systems.

I wanted to create something more accessible.

With Zearches, website owners can submit their URLs, provide a description, select a relevant category, and have valid submissions published after automated checks.

There are no mandatory submission fees or reciprocal backlinks.

The technical challenge: Trusting user-submitted URLs

One of the biggest lessons I've learned is that accepting URLs from the public introduces challenges that aren't immediately obvious.

A submission form might look simple, but behind it you need to think about:

Spam and malicious destinations
Duplicate websites and repeated submissions
URL shorteners and suspicious redirects
Misleading descriptions
Incorrect website categories
Automated bots and submission abuse
Websites that change their content after being approved

A website that appears legitimate when submitted might also change its content later.

This led me to focus heavily on submission validation, duplicate prevention, rate limits, automated checks, editorial reviews, and security monitoring.

I also learned the importance of creating backups before making changes to existing listings.

When reviewing thousands of websites, it's essential to preserve legitimate submissions while removing problematic content.

My biggest lesson: A website directory isn't just a collection of links. It's a system that needs ongoing quality control.

Even strong automated checks cannot guarantee that every external website is safe.

2. MeWebLink: Why Automation Needs Safeguards

My second project, MeWebLink, explores another approach to free website submission and discovery.

Like Zearches, it allows website owners to submit websites without mandatory listing fees or backlinks.

But working on a second directory helped me understand something important:

Automation is valuable only when you can trust how it behaves when things go wrong.

It's easy to imagine a fully automated process:

A visitor submits a URL, the system checks it, and the listing appears.

In reality, some checks can fail temporarily. A destination may redirect. A submission might resemble spam even though the website is legitimate.

Rejecting every uncertain submission can cause false positives. Automatically approving every uncertain submission creates a different problem.

A better approach is to distinguish between confidently valid submissions, clearly prohibited content, and cases requiring additional checks.

Building safer cleanup processes

Another lesson has been the importance of keeping backups and audit records before making large changes to stored data.

When maintaining an active directory, you don't want a cleanup operation to overwrite a legitimate submission that arrived after your backup was created.

I've learned to value systems that support:

Previewing changes before applying them
Backing up important data
Detecting whether records have changed
Keeping a history of editorial decisions
Recovering from unexpected results
Monitoring important website files

These safeguards make maintenance more predictable.

My biggest lesson: Good automation should make operations easier without sacrificing control, recoverability, or user trust.

3. ChipJourney: Transforming an Existing Website

ChipJourney presented a different challenge.

Rather than starting entirely from scratch, I worked on transforming an existing website into a directory while preserving its blog and useful content.

The goal was to combine two experiences: discovering submitted websites and exploring published articles.

That sounds straightforward, but changing an established website introduces different considerations.

Existing URLs, blog content, navigation, mobile layouts, internal links, and search engine indexing all need attention.

Another issue is keeping the original functionality intact while introducing new features.

Small technical decisions can have big consequences

For example, a directory submission system must preserve the intended meaning of a submitted URL.

If someone submits a particular page but the system stores only the platform's homepage, the resulting listing may not lead where visitors expect.

This kind of problem highlights how important it is to test actual user journeys rather than assuming that a successful form submission means everything is correct.

I also learned to avoid making unnecessary changes to unrelated parts of a working website.

When introducing new directory functionality, I wanted to preserve the existing blog, featured images, navigation, and important pages.

My biggest lesson: Maintaining existing functionality is just as important as developing something new.

4. What AI-Assisted Development Has Taught Me

I use tools such as ChatGPT to help develop, review, debug, and improve website functionality.

I'm not claiming to be an experienced software engineer.

My role involves defining what I want to build, understanding the problems, testing proposed solutions, deploying changes, and maintaining the websites.

AI can help translate an idea into working code, explain unfamiliar concepts, and suggest ways to troubleshoot errors.

But I've also learned that AI-generated code shouldn't be trusted automatically.

A change can appear correct while introducing an unexpected problem somewhere else.

That's why I've increasingly focused on practical safeguards before deploying changes.

My practical checklist

Make a backup first. Preserve a known working version before changing important files or data.
Test changes separately. Use a preview or dry-run mode whenever possible.
Protect existing records. Don't overwrite newer data just because an older backup contains different information.
Review security implications. Consider validation, external URLs, file access, permissions, and unexpected inputs.
Check the actual website. A successful script execution doesn't automatically mean the interface works correctly.
Keep logs. When something fails, useful records make it easier to understand what happened.
Make incremental improvements. Smaller, carefully tested updates are often safer than replacing large parts of a working website.

These steps may sound basic to experienced developers, but they've made an enormous difference in how I approach website maintenance.

5. SEO Is More Than Adding Keywords

Another area I've spent considerable time learning about is search engine optimisation.

Working on different websites has shown me that SEO involves much more than writing titles and descriptions.

Technical SEO can include:

Crawlable website pages
Consistent and descriptive URLs
Useful internal linking
XML sitemaps
Search engine indexing signals
Mobile-friendly layouts
Avoiding broken or duplicate pages
Maintaining relevant content

For website directories, there is another consideration: the quality of the content being indexed.

Automatically publishing thousands of low-quality submissions doesn't necessarily create a valuable resource.

It's important to distinguish between growing the number of listings and creating a genuinely useful website.

I've also learned that a directory backlink does not guarantee search engine rankings, indexing, or referral traffic.

My biggest lesson: Sustainable visibility starts with a website that provides genuine value to its visitors.

6. Running Multiple Projects Changes How You Think

Building three websites has helped me recognise patterns.

Different projects can share similar operational challenges even when their designs and purposes differ.

A directory needs reliable validation. A blog needs organised content and navigation. Both need dependable hosting, backups, ongoing testing, security awareness, and thoughtful user experiences.

It's also taught me the importance of prioritising.

Adding new features is exciting, but fixing an existing problem or improving reliability can sometimes deliver more value than building something new.

And perhaps the most important lesson is that launching a website is only the beginning.

Building something is one challenge. Keeping it useful, secure, and reliable is another.

7. What's Next?

I'm continuing to improve all three projects, with Zearches as my main focus.

My priorities include:

Making website discovery more useful
Strengthening automated spam and security checks
Improving website submission experiences
Maintaining better directory listing quality
Growing organic traffic and visitor engagement
Learning more about secure, maintainable web development
Exploring responsible ways to use AI-assisted development

I'm particularly interested in how independent creators can use modern tools to build useful online services without requiring large development teams.

I still have plenty to learn, and I expect there will be more mistakes, debugging sessions, and unexpected challenges ahead.

But that's part of what makes the process interesting.

Final Thoughts

You don't necessarily need a traditional programming background to start exploring ideas and building useful online projects.

However, creating something that works reliably requires patience, testing, curiosity, and the willingness to understand the tools you're using.

AI can make development more accessible, but it doesn't remove the responsibility to build carefully.

For me, the biggest achievement isn't simply having three websites online.

It's everything I've learned through trying to make them better.

Let's Discuss

I'd love to hear from the DEV community:

Have you used AI-assisted development to build or maintain a website?

What's been your biggest challenge: writing code, security, automation, deployment, SEO, or something else?

Feel free to share your experiences in the comments!

DEV Community: Zearches