Developers adopt AI tools widely but trust them cautiously in important work
Stack Overflow's latest survey of nearly 31,000 developers across 169 countries reveals that artificial intelligence tools have become deeply embedded in professional development work. However, developers remain highly sceptical about AI's reliability for critical decisions, with only 7 per cent trusting AI output for important work and nearly half (48%) only willing to trust verified results. This scepticism is particularly significant as AI agents become increasingly autonomous in the workplace.
The survey found that about 30% of developers using AI daily spend four or more hours working with it, whilst 80% of daily users engage with these tools for at least an hour daily. Two-thirds of respondents use coding assistants or agents, whilst 63% employ general-purpose chat tools. Critically, 79% of developers consider source attribution important or very important when deciding whether to trust AI-generated technical answers. This finding is particularly relevant to Stack Overflow's contentious partnerships with major AI providers like OpenAI, which faced user revolt in 2024 over training rights.
- Developers widely use AI daily but only trust it for verifiable work
- 79% demand source attribution and transparency from AI systems
- Survey reveals growing developer scepticism as AI becomes more autonomous
New here? Start with this
Artificial intelligence tools for writing and checking computer code have become widespread in software development over recent years. These tools include specialised coding assistants and general-purpose chatbots that can suggest code and answer technical questions. Developers have widely adopted them, with many now using such tools several hours a day.
Developers are nevertheless cautious about relying on AI tools for critical or important work, with research showing most would not trust AI output alone for decisions that matter. This wariness reflects concerns that AI-generated code might contain errors or outdated information, which could have serious consequences in professional use. The tension between AI's convenience and its reliability has become central to how the development community views these emerging tools.
The question matters partly because it touches on how AI companies train their systems and whether developers' own work and knowledge are properly credited. Major technology firms have struck deals with platforms like Stack Overflow, a popular website where developers share code and solutions, to use that content to train their AI. The debate over consent, compensation and attribution in such arrangements has become increasingly important to developers worldwide.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Developers are already adopting AI extensively into daily workflows, suggesting these tools deliver genuine value that outweighs abstract risks. Rather than remaining constrained by caution about worst-case scenarios, the profession should treat AI as any powerful tool requiring appropriate oversight rather than blanket distrust. As capabilities improve and experience accumulates, overly conservative approaches risk leaving the profession uncompetitive.
The case against
The survey itself vindicates developer scepticism: only 7 per cent trust AI for critical work and nearly half require verified results, reflecting documented concerns about reliability and security. Software development's real-world consequences demand rigorous verification standards, and concerns about hallucinations, vulnerabilities, and undetected errors are not theoretical but practical problems. The 79 per cent prioritising source attribution raise legitimate accountability questions that responsible developers cannot dismiss.
Read the full article at the source →
Originally published by The Register as “Stack Overflow survey finds devs hooked on AI, but not totally sold on its judgment”.