TLDR: The 2025 Stack Overflow Developer Survey indicates a significant increase in AI tool adoption among developers, with 84% now using or planning to use them. However, this widespread use is accompanied by a notable decline in trust, as nearly half of developers distrust AI output, and many report increased time spent debugging ‘almost right’ AI-generated code. Despite productivity gains for some, overall positive sentiment towards AI has decreased.
The recently released 2025 Stack Overflow Developer Survey, drawing insights from over 49,000 developers across 177 countries, highlights a paradoxical trend in the software development landscape: a surge in AI tool adoption coupled with a decline in developer trust. The comprehensive survey, now in its fifteenth year, provides a crucial snapshot of the global developer community’s evolving relationship with artificial intelligence.
According to the findings, a remarkable 84% of developers are now either using or planning to use AI tools in their development process, marking a notable increase from 76% in the previous year. Among professional developers, daily AI tool usage has reached 51%, underscoring AI’s growing integration into daily workflows.
However, this widespread adoption has not translated into confidence. The survey reveals a significant drop in trust regarding the accuracy of AI output, with only 29% of developers expressing trust, down from 40% in prior years. A substantial 46% actively distrust AI accuracy, and a mere 3% report ‘highly trusting’ the results. For experienced professional developers, this ‘high trust’ figure plummets to just 2.6%.
Developers’ frustrations are also on the rise. The number-one frustration, cited by 45% of respondents, is dealing with ‘AI solutions that are almost right, but not quite,’ which often makes debugging more time-consuming. A staggering 66% of developers report spending more time fixing such ‘almost-right’ AI-generated code, indicating that AI’s imperfections can inadvertently increase workload rather than reduce it.
When faced with untrustworthy AI answers, developers overwhelmingly prefer human collaboration, with 75% stating they would still ask another person for help. Furthermore, 35% of developers specifically turn to Stack Overflow for assistance after encountering failures with AI-generated code.
Despite these challenges, AI is playing a growing role in developer education. The survey found that 44% of developers learned new coding techniques or programming languages with the help of AI-enabled tools, an increase from 37% in 2024. Additionally, 36% of developers learned to code specifically for AI in the past year, indicating a nascent but growing investment in AI programming skills across all experience levels.
AI agents, while gaining traction, are not yet mainstream. The survey indicates that 52% of developers either do not use agents or prefer simpler AI tools, and 38% have no immediate plans to adopt them. Nevertheless, for those who do utilize AI agents, the productivity benefits are clear: 69% reported an improved workflow, and 70% experienced reduced time on specific tasks. However, AI agents’ impact on team collaboration remains limited, with only 17% believing they have improved this aspect.
Also Read:
- Large Language Models Propel Transformative Shifts Across Business, Software Development, and Automation
- Replit AI Assistant Erases Production Database, Fabricates User Data, and Misleads Developer
Overall positive sentiment towards AI tools has also seen a decline, dropping to 60% in 2025 from over 70% in 2023 and 2024. Professionals generally maintain a slightly more favorable stance (61%) compared to those learning to code (53%). Prashanth Chandrasekar, CEO of Stack Overflow, commented on the findings, stating, ‘AI is a powerful tool, but it has significant risks of misinformation or can lack complexity or relevance.’ This sentiment encapsulates the nuanced reality of AI adoption in the developer community: a powerful tool embraced for its potential, yet approached with increasing caution and a critical eye towards its current limitations.
