LLMs in Unreal Engine 5 Production: From Claude Code Promises to Measurable Reality

LLMs are rapidly entering game development workflows — from code generation to debugging and internal tools. With solutions like Claude Code, expectations are high: faster iteration, reduced costs, and leaner teams. But how do these promises hold up in real production environments?

As a technology service provider working with multiple studios across different projects and production stages, we approach LLM adoption with one key requirement: predictability. Across Unreal Engine 5 codebases — from gameplay systems to tools and pipelines — we’ve explored where LLMs provide tangible value, and where they introduce risk, inconsistency, or hidden costs.

In this talk, we share a pragmatic, field-tested perspective on integrating LLMs into UE5 development:

- Where LLMs (including Claude Code) effectively support C++ and Blueprint workflows
- Common failure patterns when applied to real production code
- The impact on code quality, maintainability, and team workflows
- Why perceived speed gains can be misleading without proper evaluation
- Practical ways to measure productivity and efficiency in LLM-assisted development

Rather than focusing on future possibilities, this session is grounded in current production realities. It aims to help studios make informed decisions about adopting LLMs — based on evidence, not expectations.