Trust in AI-generated code
Developers' reported confidence in the accuracy of AI output; trust fell from 2024 to 2025 while usage rose, and by 2026 it is mostly conditional on verification.
In Stack Overflow's surveys, trust in AI accuracy dropped from 43% (2024) to 33% (2025), with 46% actively distrusting it. The 2026 survey asked when trust is acceptable: 48% trust AI output only when they can easily verify it and 6.6% trust it with important work decisions; 76.5% run AI-generated answers locally before using them. DORA finds 30% of tech professionals have little or no trust. When they do not trust an answer, 75% go to a person. Low trust is rational given measured flaw rates, and it is why review, not generation, sets the pace.