DNXAI

How to Avoid “AI Game Slop” and Build Quality 3D Games With AI

By the DNX AI editors3 min read

I used AI to create 100 HTML games for my website, fattyberry.com, and I used AI to build the website itself. Along the way I learned why so many AI games feel like "slop", and how to make games people actually want to play.

Step 1: Turn your images into 3D models (GLB files)

To make a better game, start with good 3D assets. AI 3D tools can convert your images into GLB files, the 3D format that web games use. Some tools to try:

  • AI3D Studio, an AI 3D model generator
  • Meshy, which turns images into 3D models and exports GLB
  • Tripo, Rodin and Tencent Hunyuan3D, which also export GLB
  • TRELLIS, which works from a single image

Then prompt Claude or ChatGPT to create the game for you using those GLB files.

Step 2: Plan your game with AI first

Before you build, use Google Gemini to research games like the one you want to make, and to write image prompts for everything your game needs:

  • The icon
  • The characters
  • The road design, for example a race map

The better your plan and assets, the better your game.

Step 3: Don't build the trend; solve a problem

This is where most people go wrong. When everyone builds the same trending game, it becomes slop. Instead:

  • Look for games with bad reviews and build the solution.
  • Understand your players. Ask AI how players behave, what they enjoy and why they quit.
  • Build for your own people. Local stories and everyday life are a huge opening.

A real example: Nigerian developer Shalom Mathew built Lagos Life, a life simulation game set in Lagos and inspired by The Sims. She reportedly built it in under 12 hours, and it reached 1 million players in three days. It was inspired by Lagos Run, a game by Opeyemi Adeniran where you drive a danfo bus through Lagos traffic. Lagos Life is now becoming a gaming company. (Read the story on Condia)

That game didn't win by copying a trend. It won because millions of people saw their own life in it.

Rebuild games to learn, but don't copy to publish

You can rebuild a game you like just to learn how it works: how it functions, and which prompts make it better. That's a great way to study.

But don't publish a copy. The game industry is booming, and copyright problems can make all your hard work and the money you spent on AI go to waste. Take your time, learn, gather your information, and then build your own better game.

Play your own game

You have to enjoy the game yourself. Play it like a player would:

  • Is it smooth?
  • Is it fun after 5 minutes?
  • How can I make this game better?

Then ask AI those same questions and keep improving.

Change your prompts: understand what you build

One day my prompts changed. I kept hearing people say vibe coders aren't real programmers. So now I write prompts that help me understand what I'm building, not just get it done, and I use data and information to guide every game.

Don't just get a reply, install the game, and say "I'm done". Don't build in 3 minutes a game that took big companies huge money to make and call yourself a hero. It's time to understand the market and the information you need to build a better game.

Get your game seen

A great game nobody finds makes no money. Connect your game to search and social:

  • Search: give every game its own page with a clear title and description, so Google can find it.
  • Social: post short gameplay clips (5–15 seconds) on TikTok, YouTube Shorts and Reels.
  • Community: share your game where players hang out, and listen to their feedback.

In my next post, I'll show you how to use AI the right way.

More guides