A Stack Overflow pulse survey of ~900 developers (conducted in partnership with OpenAI) reveals that AI use for learning has surged to 64%, up from 37% in 2024, while the number of learning resources developers rely on has sharply declined. Despite growing adoption, trust remains the top barrier—38% cite distrust in AI results—and experienced developers still favor technical documentation as a first step. Most developers use AI alongside other resources (technical docs, search, Stack Overflow) rather than exclusively. The survey also explores the 'AI productivity tax': the hidden cost of verifying AI outputs and the lack of citation provenance in LLM answers. On AI job platforms, developer appetite is lukewarm—only 17% would use an AI agent for job search unconditionally, with human oversight and data transparency being key conditions for broader acceptance.
Table of contents
Is it still learning if it is easier than it used to be?AI’s productivity taxWhen is a human required?Sort: