We’re in the middle of a developer renaissance powered by AI — yet a large portion of the community remains cautious, especially when it comes to the next frontier: Agentic AI. The just‑released 2025 Stack Overflow Developer Survey paints this nuanced portrait vividly: 84% of respondents use or plan to use AI tools in their workflows, yet less than a third trust their accuracy — just 29% this year, down from much higher levels in prior surveys.
A Tale of Two AI Worlds: Generative vs. Agentic
On one hand, generative AI tools — ChatGPT, GitHub Copilot and others — have become near-ubiquitous. OpenAI’s GPT models are used by 81–82% of developers, followed by Claude Sonnet (~43–45%) and Gemini Flash (~35%). And more than half of professional developers (50.6%) use AI tools daily. It’s mainstream.
But agentic AI — AI that takes multiple-step, autonomous actions — remains emergent. Only 31% are currently using AI agents, 17% plan to, and a significant 38% plan not to use them at all. That aligns with the Reddit‑driven takeaway that nearly 40% aren’t on board with agents — at least for now.
Is that believable? Yes — and instructive. Developers are pragmatists. They’ll use tools that reliably boost productivity — like autocomplete, code suggestions and troubleshooting helpers — but they’re not yet ready to delegate multi-step tasks to AI agents. The maturity, transparency and accuracy simply aren’t there yet.
Warming Up to AI — But with Side-Eyed Caution
The broad adoption of generative AI is undeniable, but trust is eroding. Favorable sentiment toward AI tools overall has dropped from over 70% in prior years to around 60% in 2025. About 46% of developers say they don’t trust the accuracy of AI outputs, up sharply from 31% last year.
Why the distrust? “AI outputs are almost right — but not quite,” says 45% of developers; 66% spend more time debugging AI-generated code than expected. And when things get complex, they still turn to humans: 75% report they’ll still ask another person for help when they don’t trust AI answers.
This is the developer four‑step pattern with AI: Use it, verify it, understand it and only delegate if it’s proven.
Claude Leads (But Narrowly), ChatGPT Reigns
Stack Overflow’s survey shows Claude Sonnet is the most admired AI model, though OpenAI’s GPT remains the most used. Specifically, Claude garners admiration (61.2%) vs GPT (67.5%) and is the second most desired. For usage, GPT leads (~82%), Claude Sonnet is in the mid‑40s, and Gemini Flash is around 35%.
The takeaway? Developers are exploring beyond ChatGPT even as it holds the share lead, but admiration and curiosity swirl around newer players like Claude.
The State of Transition
All the signs point to a community in transition. Generative AI is widely adopted and embedded across workflows, from writing code to learning new frameworks. Yet agentic AI remains cautious territory, held back by trust, reliability and real-world complexity concerns. Sentiment toward AI has cooled, trust has declined and developers are consciously balancing automation with oversight.
But this is not pessimism — it’s maturity. The script is no longer “AI will save us all.” The script is “AI helps us, but it’s still a tool, not a trusted partner — yet.”
Final Word: Developers in Flux, AI Accelerating Wisely
The 2025 Stack Overflow survey tells a compelling story: Developers are rapidly adopting AI, favoring trusted generative models while keeping agentic AI at arm’s length. Satisfaction may have dropped, but usage remains high. Favorites like ChatGPT dominate, but admiration for Claude signals emerging shifts in preference.
We’re not at the agentic AI tipping point — yet. But the foundations are being laid. As tools mature, trust evolves and real-world utility proves itself, developers will increasingly invite agentic AI into their toolchains — but on their terms, with oversight and only when it earns its keep.
That’s not hesitancy. That’s sensibility. It’s the hallmark of an ecosystem moving smartly, not blindly, toward the future.
