I have been writing code for 13 years — primarily Unity, C#, XR development, and more recently web work with HTML, CSS, and JavaScript. AI coding tools have become part of my workflow over the last couple of years, and my experience with them is specific to how I actually work rather than a general survey of what the internet says is best.
I will give you my honest assessment of what I have used and found useful. Where I have heard credible things from other developers — including my cousins who work in different tech stacks — I will say that too and be clear it is not my firsthand experience.
What I Actually Use
🤖 ChatGPT — Two Years as My Primary Coding Assistant
For about two years, ChatGPT was the AI tool I reached for most in my development work. For Unity and C# specifically — component logic, shader code, debugging error messages, generating boilerplate that would take time to write from scratch — it was genuinely useful. Not perfect, but useful. It would get 70-80% of the way there on most tasks and I would clean up the rest. I got comfortable with how it responds, learned how to break complex requests into manageable chunks, and knew when to push back on a wrong answer.
The way I used it most in Unity work was for specific, bounded problems. "This component needs to track the position of a target object and smooth its movement — write me a lerp-based tracking script." Or debugging: paste in an error message, describe the context, get a fix. For that kind of focused, isolated task, ChatGPT was reliable over two years of daily use.
Where I hit limits was in web development. When I built a project using HTML, CSS and Three.js, ChatGPT kept giving me code that worked in isolation but did not connect properly when the project got complex. It would suggest approaches that conflicted with how I had already structured things. That frustration is what pushed me to try Claude.
🧠 Claude — Better for Web and Understanding Your Existing Code
I switched to Claude as my primary AI tool after that Three.js experience. I described the integration issues to Claude, pasted in my existing code, and it read through everything before suggesting a solution. It understood what I had already built and gave me code that actually fitted my existing structure rather than assuming I was starting from scratch.
That is the core difference I have experienced between Claude and ChatGPT for coding: Claude is better at understanding context you give it and working within it, rather than giving you a generic solution that you then have to adapt. For web development — HTML, CSS, JavaScript, Three.js — this makes a significant practical difference. For Unity and C# work I would say they are closer in quality, with ChatGPT having the edge from familiarity and two years of workflow habit.
Claude's context window is also genuinely useful for larger projects. You can paste in multiple files or a long conversation history and it keeps track of it all. With ChatGPT I found myself re-explaining context I had already covered earlier in the session. Claude does this much less.
⌨️ Cursor — Tried It Recently, Promising for Other Stacks
I tried Cursor recently — it is built on VS Code which is one of my main editors, so getting started was straightforward. For Unity work specifically, my honest assessment is that it is okay. The Composer feature — where you describe a change and it edits across multiple files — is interesting in concept, but Unity projects have a specific structure that does not always play nicely with how Cursor approaches multi-file editing.
Where I expect Cursor really shines is in web and full-stack development. My cousins who work in React, Node, and Python tell me Cursor has significantly changed how they work — the codebase chat feature especially, where you can ask questions about your entire repository and get relevant answers. For that kind of development, where the AI can understand the full project context and make changes across a coherent file structure, I can see why it would be transformative. For Unity XR development, the gains are more modest in my experience so far.
🔧 GitHub Copilot — Solid Autocomplete, Similar Story
I also tried GitHub Copilot recently, using it in both VS Code and Visual Studio — both of which I use depending on the project. The autocomplete is fast and the suggestions are generally sensible for Unity C# work. It integrates cleanly into both editors without disrupting your existing workflow, which is its biggest practical advantage over tools that require you to change how you work.
The experience was similar to Cursor — good, but not dramatically better than what I was already getting from using ChatGPT and Claude via chat for specific questions. Again, developers I know who work in other technologies — web, Python, TypeScript — consistently report bigger gains from Copilot than I have seen in Unity work. The tool seems to perform better when the codebase structure is cleaner and more standardised than a typical Unity project.
Tools I Have Not Used But Are Worth Knowing
🚀 Bolt.new — Full App from a Prompt
Bolt.new generates a working full-stack web application from a text description — frontend, backend, database, deployed and live. I have not used it myself but the concept is interesting for rapid prototyping and MVPs. For someone who needs a working web app quickly without deep development knowledge, it is worth exploring. For experienced developers working on complex production systems, it is more of a starting point than a finished product.
▲ v0 by Vercel — React UI Generation
v0 generates production-ready React component code from a description — using shadcn/ui and Tailwind CSS. The output quality is reportedly very high, looking like something a senior frontend developer wrote. I work primarily in Unity rather than React so I have not had a use case for it personally, but for frontend developers building UI components it seems genuinely useful.
🆓 Codeium — Best Completely Free Option
Codeium is free for individual developers indefinitely, supports 70+ languages, and integrates with all major IDEs including VS Code and Visual Studio. If you want AI autocomplete without any subscription cost, this is the starting point. The quality is reportedly close to Copilot in 2026, though I have not tested it personally.
My Honest Verdict for Unity and XR Developers
If you are a Unity developer specifically, my current recommendation based on what I have actually used is: Claude via chat for web work and complex problems that need context understanding, ChatGPT for Unity-specific C# tasks where two years of familiarity with how it responds to Unity problems still gives it an edge for me personally, and Cursor or Copilot as a VS Code/Visual Studio companion for the autocomplete and in-editor convenience.
The gains from Cursor and Copilot in Unity work are real but modest compared to what developers working in React, Python, or TypeScript seem to experience. If your stack is Unity-heavy, I would not expect the same transformation your web developer colleagues might be seeing. That could change as these tools improve — Unity support has been getting better — but that is where things stand from my experience in early 2026.
Quick Decision Guide
Frequently Asked Questions
From my own experience: ChatGPT for Unity-specific C# tasks where I have two years of workflow familiarity, and Claude for anything requiring understanding of existing code structure. Cursor and Copilot work in VS Code and Visual Studio but the gains are more modest for Unity work than for web or Python development in my experience.
It is okay for Unity — the VS Code foundation means setup is easy and familiar. But the biggest benefits people report from Cursor seem to be in web and full-stack development where the codebase structure is cleaner and more standardised. Unity projects have some quirks that limit how much Cursor's multi-file editing features can help. Worth trying on the free tier to form your own opinion.
For Unity and C# — ChatGPT, based on two years of familiarity with how it responds to Unity-specific problems. For web development and anything requiring deep understanding of existing code structure — Claude is clearly better in my experience. The Three.js project that pushed me to switch is the clearest example: Claude read my existing code and gave me solutions that fitted it. ChatGPT kept giving me generic solutions I had to adapt.
For web and full-stack developers — most reports say yes, the time saved on autocomplete and boilerplate covers the cost easily. For Unity developers specifically, my experience is that the gains are more modest. The free tier now available with a GitHub account is worth trying before committing to a paid plan.