Article illustration 1

Main article image showing a developer interacting with AI tools. Credit: fotograzia/Getty Images

Developers are caught in an AI contradiction: They're using artificial intelligence more than ever but trusting it less. That's the stark revelation from Stack Overflow's 2025 Developer Survey, which polled nearly 50,000 developers globally. Adoption has surged—84% now use or plan to use AI tools (up from 76% in 2024), with 51% of professionals relying on them daily. Yet sentiment has cratered: Only 60% express positivity toward AI, down from over 70% in prior years. Even more alarming, just 3% "highly trust" AI outputs—a figure that drops to 2.6% among seasoned developers.

The Collapse of Confidence

Trust isn't merely stagnating; it's collapsing. In 2024, 43% felt good about AI accuracy. By 2025, 46% actively distrusted outputs, while only 33% expressed any trust. This erosion stems from tangible frustrations:
- 66% cited "AI solutions that are almost right, but not quite" as their top pain point
- 45% reported "debugging AI-generated code is more time-consuming"

As Bill Harding, CEO of Amplenote and GitClear, bluntly stated in his analysis of 211 million lines of code:

"Developers trust the current generation of AI assistants about as much as we trusted the previous generation, i.e., not much."

Real-world disasters like Jason Lemkin's Vibe project—where an AI agent deleted a production database—highlight why skepticism runs deep. This distrust extends beyond developers: Only 8.5% of Americans "always trust" Google's AI Overviews, per recent data, while a KPMG study found just 46% of people globally trust AI systems.

The Technical Debt Time Bomb

Rushed AI adoption risks long-term fallout. Junior developers—who trust AI most—risk skill atrophy. Independent developer Namanyay Goel warns:

"We're trading deep understanding for quick fixes, and while it feels great in the moment, we're going to pay for this later."

GitClear's research confirms a direct link between AI adoption and rising defect rates. Harding cautions that measuring productivity by lines of code or commits exacerbates this:

"Leaders need to recognize that more code is often worse. Copying and pasting code leads to higher defect rates."

Tools, Languages, and the Human Imperative

Despite AI's rise, traditional tools dominate:
- Visual Studio (75%) and VS Code (29%) remain top IDEs
- Vim and Notepad++ persist as favorites

Python's popularity soars (driven by AI libraries like TensorFlow), but Rust (83% approval) reigns as the most admired language. OpenAI's GPT models dominate LLM usage (82%), followed by Anthropic's Claude and Google's Gemini.

Crucially, 75% of developers insist human advice is irreplaceable for critical tasks where AI falters. And AI agents remain niche: 38% have no plans to adopt them soon.

This isn't just about tools—it's about the soul of craftsmanship. As AI generates brittle, error-prone code, the survey underscores that developer intuition and oversight aren't relics; they're the bedrock of reliable software. The industry must redefine productivity beyond raw output, or risk drowning in AI-amplified technical debt.

Source: Based on the 2025 Stack Overflow Developer Survey and reporting by Steven Vaughan-Nichols for ZDNet.