What shapes developers’ trust in AI coding tools?

What they studied

The authors studied what shapes developers’ trust in generative AI tools and their intentions to use those tools at work. Their survey included 238 participants at GitHub and Microsoft.

The study examined both tool characteristics and individual differences. These differences included risk tolerance, enthusiasm for technology, and confidence in using computers. The authors used statistical modeling to investigate relationships between these factors, trust, adoption intentions, and reported usage. They also analyzed developers’ descriptions of challenges and risks when using generative AI.

What they found

Several aspects of the tools were significantly associated with trust: system and output quality, functional value, and the ability to stay aligned with task goals. Examples included presentation, safety and security, performance, and educational and practical benefits.

Trust was, in turn, significantly associated with developers’ intentions to adopt the tools. Developers’ cognitive styles were also associated with those intentions. Adoption intentions were then associated with reported usage.

A separate analysis identified factors that were important but insufficiently supported by existing tools, suggesting targets for design improvements. However, the abstract does not name those specific under-supported factors or provide a ranked list. The qualitative analysis helped explain why gaps persisted in development contexts, though the abstract does not detail those explanations.

These are survey-based associations, not evidence that changing any single feature will necessarily increase trust or usage.

Why it matters for builders

For beginners choosing an AI coding tool, the findings suggest looking beyond whether it can produce code. Useful evaluation questions include whether its output is understandable, whether it provides practical or educational value, and whether it stays focused on the task you actually asked it to complete.

The study also suggests that one workflow may not suit everyone. Your comfort with risk and confidence with computers are relevant considerations when deciding how to integrate AI assistance. The research does not establish a best tool or workflow; it highlights dimensions worth considering.

Source: "What Needs Attention? Prioritizing Drivers of Developers’ Trust and Adoption of Generative AI" (2026-08-22, ACM Transactions on Software Engineering and Methodology). This is our summary of the paper's abstract — read the original for full details.