Blog · AI and teams
Almost half of developers distrust what AI writes, and 84% use it or plan to
In Stack Overflow's 2025 survey, 46% of developers say they distrust the accuracy of AI tools and only 33% trust it, even though 84% already use these tools or plan to. The most experienced developers are the most skeptical, and 87% have concerns about the accuracy of agents. What each team still has to settle is who reviews what AI produces.
What the survey says
Usage grew: 84% use or plan to use AI tools for coding, up from 76% in 2024, and 50.6% of professional developers use them every day. Trust moved the other way. Only 3.1% highly trust the accuracy of what these tools produce, and favorable sentiment fell from over 70% in 2023 and 2024 to 60% in 2025.
Developers with more experience are the most cautious: 20.7% highly distrust the accuracy. With agents, the doubts run higher: 87% have concerns about their accuracy, and 81% about data security and privacy. And 52% don't use agents or stick to simpler features, such as autocomplete.
Where is the real cost?
In what almost works. The most cited frustration, named by 66%, is AI solutions that are almost right, but not quite. And 45.2% say debugging AI-generated code takes them more time. An obvious error gets thrown out quickly; a subtle one gets past the first read and shows up later.
When they don't trust an answer, 75% still ask a person. The survey shows that human verification is still needed at every experience level. AI has become part of daily work, but it doesn't replace the judgment of the person reviewing it.