Article illustration 1

Main image: Getty Images

Programmers are caught in an AI paradox. According to Stack Overflow’s 2025 Developer Survey—which polled nearly 50,000 developers—84% now use or plan to use AI tools in their workflows, a significant jump from 76% just a year ago. Over half (51%) of professional developers rely on them daily. Yet beneath this adoption surge lies deepening skepticism: Only 60% expressed positive sentiment toward AI, down from over 70% in both 2023 and 2024.

The Trust Collapse

Trust metrics reveal an even starker reality. In 2024, 43% of developers felt confident in AI accuracy. By 2025, only 33% trusted outputs, while 46% actively distrusted them. Among seasoned professionals, just 2.6% reported high trust, with 20% expressing strong skepticism. This erosion follows high-profile failures like Jason Lemkin’s Vibe project, where an AI agent wiped a production database.

Why the Disconnect?

Two frustrations dominate:
1. 66% battle "AI solutions that are almost right, but not quite"
2. 45% find debugging AI-generated code more time-consuming than writing it from scratch

Bill Harding, CEO of Amplenote and GitClear, contextualizes this 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." His research links rising defect rates directly to AI adoption.

Article illustration 2

Coding concept (Getty Images)

The Technical Debt Time Bomb

Junior developers—the most trusting demographic—face particular risk. Independent developer Namanyay Goel warns: "We're trading deep understanding for quick fixes... we're going to pay for this later." Harding echoes this, cautioning that if companies measure productivity by lines of code written, "AI-driven technical debt will spiral out of control." GitClear’s data shows copied AI code correlates with higher defect rates.

Adoption Amid Mistrust

Despite skepticism, OpenAI’s GPT models dominate (used by 82% of AI-adopting developers), followed by Anthropic’s Claude and Google’s Gemini. Traditional tools remain preferred: Visual Studio (75%) and VS Code (29%) outpace AI-first IDEs, while Vim and Notepad++ persist—proving that efficiency doesn’t override caution.

The Irreplaceable Human

Looking ahead, 75% of developers insist human expertise remains essential when AI outputs seem unreliable. And AI agents? Over half still use simpler tools, with 38% having no plans to adopt autonomous coding agents soon. As Stack Overflow’s data scientist notes: "Trust isn’t binary—it’s earned through precision. When systems hallucinate solutions or introduce subtle bugs, skepticism becomes professional survival."

Source: Stack Overflow 2025 Developer Survey, GitClear AI Copilot Code Quality Study, ZDNet analysis