Stack Overflow 2025 Survey: AI Adoption Reaches 79% Yet Trust Declines Sharply

Stack Overflow's review of its 2024 and 2025 Developer Survey data shows artificial intelligence has become a pervasive part of workflows, but higher usage has not brought a corresponding rise in enthusiasm or trust.

From Coding Assistants to AI Agents

The 2025 survey measured AI agent usage for the first time: 31% of respondents reported using them. ChatGPT and GitHub Copilot remain the primary entry points, used by 82% and 68% of those who use off‑the‑shelf assistants, followed by Google Gemini (47%), Claude Code (41%), and Microsoft Copilot (31%). A smaller April 2026 user survey found AI agent adoption rising to roughly 59%, with Claude Code climbing to 55%. The survey notes that agent use correlates with daily users and those in execution‑focused roles, but agents are not yet fully autonomous in software engineering; the most common uses remain exploration, drafting, and assistance.

In 2024, 82% used AI to write code, 68% for search, 57% for debugging, and 40% for documentation. Direct comparison with 2025 is limited due to question changes, but the overall trend indicates coding is still largely AI‑assisted rather than fully delegated to AI.

Learners Use More Tools But Trust Them Less

Among programming learners, AI tool usage rose from 63.0% (2024) to 73.3% (2025), with 39.5% using them daily. Meanwhile, positive sentiment fell from 72.2% to 52.8%, while skepticism rose from 6.4% to 26.3%. The share of students who moderately or highly trust AI output accuracy dropped from 49.1% to 37.4%, and those positive about AI for complex tasks fell from 41.5% to 33.7%. The analysis highlights a key distinction: generating code that runs is not the same as understanding why it runs. AI agents have not gained broad acceptance among students; 28.4% had tried them, while 44.1% said they do not intend to adopt them.

Work Environment Grows More Complex

Fully remote work declined from 38.0% to 32.4%; fully on‑site stood at 17.9%, organized hybrid at 37.1%, and a flexible “choose when to come in” category at 12.6%. Overall, 82% had some remote flexibility in 2025 versus 80% in 2024. Geographic differences: the U.S. led in fully remote (45.0%), India in fully on‑site (30.5%), while Germany showed 20.7% in flexible on‑site and 24.6% in flexibility‑oriented hybrid. Job satisfaction split: 28.4% dissatisfied, 47.1% neutral, 24.5% satisfied. Top satisfaction drivers were autonomy and trust in task management, followed by compensation/benefits and solving real problems. 85.3% of respondents were individual contributors, 14.7% people managers.

What This Means for Developers

The data paints a more complex picture than a simple “pro‑AI vs. anti‑AI” binary. AI adoption is expanding rapidly, yet developers continue to scrutinize output quality and limitations, and still seek autonomy and human context. When asked about a future where systems handle most programming tasks, 61.3% said they would turn to other people to deeply understand something, and 58.1% would do so to learn best practices. The survey measures changes inside work and learning, not just tool penetration. The real shift is AI moving from an optional experience to a pervasive work layer, accompanied by tighter scrutiny of accuracy, understanding, and professional responsibility. Year‑over‑year comparisons are imperfect due to question and sample differences; the relationship between AI use and learning improvement remains an open question. The 2026 Developer Survey will show whether the usage‑satisfaction gap persists and whether agent adoption truly leads to more delegation of decisions and tasks.

Reference: https://stackoverflow.blog/2026/10/01/a-look-back-before-we-look-forward-a-developer-survey-retrospective/