Increasing use of AI among developers leads to declining trust
Takeaways from the latest Stack Overflow survey
The latest developer surveys reveal a fascinating paradox: while the use of AI tools in software development is growing rapidly, trust in these technologies is declining at the same time.
According to the Stack Overflow Developer Survey 2025, which surveyed over 49,000 participants, 80% of developers now use AI tools in their workflows—a significant increase over previous years. Paradoxically, however, confidence in AI's accuracy has fallen from 40% to just 29%, while overall approval of AI tools has declined from 72% to 60%.
The main reason for this development lies in a specific problem: 45% of respondents cited “AI solutions that are almost right, but not quite” as their biggest frustration. These “almost right” results are particularly problematic because they can introduce bugs that are difficult to identify and require time-consuming debugging. 66% of developers report that they spend more time correcting AI-generated code.
As Ars Technica reports, this problem particularly affects junior developers, who can develop a false sense of security through their reliance on AI tools.
Despite widespread use, 75% of developers say they would still turn to other people when uncertain, rather than relying solely on AI responses. This underscores the importance of human expertise and trusted developer communities in an increasingly AI-powered work environment.
The survey makes it clear that the future of software development will be shaped not only by powerful tools but above all by trust, reliability, and the integration of human expertise.