Stack Overflow survey finds developers wary of AI tools

Stack Overflow has published its 2026 Developer Survey, based on more than 30,000 responses from developers in 169 countries.

The findings show that artificial intelligence is now widely used by developers, but trust in its output remains limited and often depends on whether users can verify the results themselves.

Nearly half of respondents said they trust AI output when they can verify it, while 78.7% said source attribution matters when judging AI-generated code or technical answers. Only 6.6% said they would trust AI with an important decision.

Coding assistants or agents were used by 65.9% of respondents, and 62.5% used general-purpose chat tools. Automated agent workflows were used by 26.2%, while 17.2% said they did not use AI tools.

Use was strongest in tasks closest to software production. Some 74.2% said AI was useful or very useful for implementation, and 72.2% said it was useful for debugging or refactoring code.

That dropped sharply for less technical work. Around 34.4% of respondents said AI was useful for tasks such as team communication or design.

Organisational rules have not kept pace with adoption. Some 29.7% of developers said AI use at work was optional and left to the individual, while only 24% said their employer had approved tools or an official AI policy or guidance.

Prashanth Chandrasekar, Chief Executive Officer of Stack Overflow, said developers were applying caution as AI systems took on more work.

"Developers are historically a highly skeptical group, and over the last year, as AI agents become rapidly more autonomous, that skepticism now acts as a critical safeguard," Chandrasekar said.

"As agents take on more work, developers need to know where an answer came from, what context informed the information that was produced or a decision that was made, and, most crucially, that they can actually verify its accuracy and relevance. Trust is not a given when it comes to developers; it is earned through transparency, source attribution, and rich technical context."

Learning habits

The survey also pointed to continued demand for new skills. Some 52% of developers said they had learned a new coding skill or programming language over the past year.

Among those learners, 58.9% used technical documentation, 52.7% used AI code-generation tools, 51.7% used other online resources, and 33.8% turned to Stack Overflow.

When developers hit a technical problem, online search remained the most common starting point at 82.7%. AI followed at 69.9%, ahead of maintained repositories at 43.7% and directly contacting a trusted person at 28.8%.

The data also suggested that heavy AI users are spending significant time with the tools. Among daily AI users, 80.7% said they spent at least an hour a day with AI, and 31.1% said they used it for four hours or more.

Workplace shifts

Beyond AI, the survey described a workforce facing mixed conditions. It found that 45.1% of developers were complacent in their current jobs, compared with 32.6% who were unhappy and 22.3% who were happy.

Among those classed as complacent, burnout or tech fatigue was the leading cause at 21.1%. Unclear priorities were the most commonly cited source of workplace difficulty at 16.6%, followed by inefficient workflows at 14.6% and insufficient time or resources at 14.1%.

Remote work remained the most common arrangement, accounting for 34.8% of respondents. Hybrid arrangements leaning in-person made up 19.9%, while 17.6% said they worked fully in person.

Self-employment also rose. Freelancers and sole proprietors accounted for 10.5% of respondents, up from 3.9% the previous year.

Technology choices

When recommending technology, developers continued to prioritise practical considerations. Reliability, performance, and stability were the leading reasons at 17.3%, ahead of productivity or time savings at 15.9% and community or ecosystem strength at 12.9%.

AI features were cited by 2.3%, placing them well below those other factors.

Established programming tools remained dominant in the results. JavaScript led for the twelfth consecutive year at 62%, followed by SQL at 58.4%, HTML and CSS at 58.2%, and Python at 58%.

In databases, PostgreSQL was used by 57.8% over the past year, compared with 42.1% for SQLite and 35.1% for MySQL.

The survey also found close usage levels among large language model providers. Anthropic was used by 74.4% of respondents and OpenAI by 74.1%, but future preference tilted more heavily towards Anthropic, with 49% saying they wanted to use it compared with 32.1% for OpenAI.

Across software work, 77.3% said they built small scripts or automations, 63% built team-level internal tools, and 54.4% worked on customer-facing products. Meanwhile, 18.6% said they built software intended to replace paid vendors.