Imagine asking an AI to write code for you. Sounds amazing, right? But how can you be sure it's correct, especially if you don't understand how the AI "thinks"? This is a core problem in AI for code today: a lack of trust and interpretability. New research introduces a clever method called ASTrust to address this. Instead of just measuring if AI-generated code *works*, ASTrust explains *why* the AI made certain coding choices. It does this by linking the AI's confidence level to the actual structure of the code, using something called an Abstract Syntax Tree (AST). Think of an AST like a blueprint of your code's grammar. ASTrust overlays the AI's confidence onto this blueprint. So, you can see exactly which parts of the generated code the AI is sure about, and where it might have made a mistake. Researchers tested ASTrust with twelve different large language models (LLMs) on code from popular GitHub repositories. They also conducted a human study to see how developers perceived the explanations. The findings? ASTrust makes it much easier to spot potential errors and understand the reasoning behind the AI's code suggestions. The implications are big. With better explanations, we can gain more confidence in using AI coding tools. ASTrust also helps researchers understand what types of coding structures are easier or harder for AIs to learn, potentially leading to more robust and reliable AI coding assistants. This represents a big step towards more trustworthy AI-generated code, but it's not the end. The next big questions revolve around expanding the analysis to even larger models and more diverse programming languages. Making explanations even more user-friendly is also key to wider industry adoption of AI-powered coding tools.