THE STATUS OF ARTIFICIAL INTELLIGENCE IN THE DEVELOPER COMMUNITY: A LITERATURE REVIEW OF STATISTICS, TRENDS, AND EXPECTATIONS CURRENT ADOPTION PATTERNS, EMERGING TRENDS, AND FUTURE OUTLOOK (2023–2026)

Abstract

Artificial intelligence (AI) has moved from an experimental novelty to a structural component of professional software development within roughly three years. This literature review synthesizes evidence from four large-sample industry surveys: the Stack Overflow. (2025) (n = 49,009), the Google Cloud. (2025) (n > 5,000), the JetBrains State of the Developer Ecosystem 2025 (n = 24,534), and Github. (2025) platform telemetry. This study presents a targeted set of peer-reviewed and preprint academic studies, including five randomized controlled trials, to assess the current status, dominant trends, and near-term outlook of AI adoption in the global developer community. Convergent survey evidence indicates that 84% of developers now use or plan to use AI tools (up from 76% in 2024), with 51% of professional developers using AI daily and adoption reaching 90% in DORA’s sample. Despite this, trust in AI accuracy has fallen to 29–33%, down from roughly 40% in prior years, and positive favorability has declined from 72% to 60%, producing a well-documented adoption–trust paradox. Unlike prior industry-report syntheses, this review foregrounds contradictory causal evidence: while an early randomized controlled trial found AI assistance produced a 55.8% speed gain on a scoped, unfamiliar task, the most methodologically rigorous field experiment to date, a 2025 trial by METR, found that AI access slowed experienced open-source developers by 19% on real maintenance work, even though those same developers believed AI had made them faster both before and after the study. Additional peer-reviewed evidence links AI-assisted coding to reduced code security, rising code churn and duplication, and measurable deficits in skill formation among less experienced programmers. The review concludes that AI adoption is now unambiguous and structural, but that its net effect on software quality, security, and the profession’s skill base remains contested, and appears to depend more on task complexity, codebase familiarity, and developer experience than on the technology in isolation.