Stack Overflow survey of ~90K developers in 185 countries: 70% use or plan to use AI coding tools in 2023; only 3% “highly trust” and 39% “somewhat trust” them
James Vincent / The Verge :
Context & Ripple Effects
Stack Overflow's annual survey has long been the industry's baseline read on developer sentiment — its 2018 edition found 73% of respondents more excited by AI than concerned about its dangers (Rust, Kotlin, and Python topped that most-loved list). The 2023 edition marks the pivot point: with ~90,000 developers across 185 countries reporting, adoption of AI coding tools has become mainstream behavior while confidence in them has not.
First-order effects
- Seven in ten developers are now using or planning to use AI coding assistants in their 2023 workflow, but the 3%-highly-trust figure means the output gets treated as a draft to verify, not a result to ship — the review burden shifts onto every developer who adopts.
Second-order effects
- Tool vendors compete on accuracy rather than capability, since the trust deficit is the binding constraint; and the Wired survey of 730 developers shows over half of employers already paying for AI tools, so budget holders are funding a category their own staff rate as unreliable.
Third-order effects
- The gap persists into the follow-up data: by the 2025 Stack Overflow survey, usage climbs to 84% but trust in AI accuracy falls from 43% to 33% — suggesting distrust is not a maturity phase that resolves, but a structural condition of AI-assisted development that shapes tooling, review practices, and hiring expectations.
The trend: Developer adoption of AI coding tools keeps climbing faster than trust in them does, making verification-heavy workflows the default state of software engineering.