Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Quo Vadis, Code Review? Exploring the Future of Code Review

Michael Dorner, Andreas Bauer, Darja Šmite, Lukas Thode, Daniel Mendez, Ricardo Britto, Stephan Lukasczyk, Ehsan Zabardast, Michael Kormann

TL;DR

The paper investigates the future of code review by surveying 100 professional developers across five organizations. It finds that code review will remain essential with similar or greater time commitment and a broader range of artifacts reviewed, while LLMs are expected to become active participants. It also identifies three erosion narratives—understanding, accountability, and trust—that could arise from AI involvement, arguing for sustained human oversight to preserve interpretability and governance in software engineering. The work highlights practical implications for managing AI integration in code review and emphasizes maintaining human-centric accountability amid increasing automation.

Abstract

Code review has long been a core practice in collaborative software engineering, yet its future trajectory is unclear. In this research, we examine how professional developers experience code review today and what changes they anticipate in the next five years. We conducted a survey with 100 developers from five software-driven companies, capturing current review effort, reviewed artifacts, and expectations about future practice. Practitioners expect code review to remain essential, with similar or greater effort and a broader range of artifacts under review. At the same time, almost all expect LLMs to become active participants in code review. With this new participant in code review, we see long-term risks of eroding human understanding, accountability, and trust. Code review may therefore act as a lens through which the challenges of AI in software engineering become visible first.

Quo Vadis, Code Review? Exploring the Future of Code Review

TL;DR

The paper investigates the future of code review by surveying 100 professional developers across five organizations. It finds that code review will remain essential with similar or greater time commitment and a broader range of artifacts reviewed, while LLMs are expected to become active participants. It also identifies three erosion narratives—understanding, accountability, and trust—that could arise from AI involvement, arguing for sustained human oversight to preserve interpretability and governance in software engineering. The work highlights practical implications for managing AI integration in code review and emphasizes maintaining human-centric accountability amid increasing automation.

Abstract

Code review has long been a core practice in collaborative software engineering, yet its future trajectory is unclear. In this research, we examine how professional developers experience code review today and what changes they anticipate in the next five years. We conducted a survey with 100 developers from five software-driven companies, capturing current review effort, reviewed artifacts, and expectations about future practice. Practitioners expect code review to remain essential, with similar or greater effort and a broader range of artifacts under review. At the same time, almost all expect LLMs to become active participants in code review. With this new participant in code review, we see long-term risks of eroding human understanding, accountability, and trust. Code review may therefore act as a lens through which the challenges of AI in software engineering become visible first.

Paper Structure

This paper contains 17 sections, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Method
  3. Study Design
  4. Context and Sample
  5. Survey Instrument
  6. Data Collection
  7. Data Analysis
  8. Data Availability
  9. Code Review Five Years From Now
  10. Not Less Time
  11. Broader Coverage
  12. The New Code Review Participant
  13. Doesn’t LGTM: Implications for the Future of Code Review
  14. Erosion of Understanding
  15. Erosion of Accountability
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Cumulative distribution of hours spent on code review today ($\leq 10$ hours). In the absence of raw data, we approximated the cumulative distribution from the reported ordinal-scale results at Microsoft Bosu2017.
  • Figure 2: Practitioners anticipate reviewing a broader range of artifacts in the future, with notable growth in attention to GUI-based test code and a decline in those reviewing no artifacts at all.
  • Figure 3: Practitioners expect the roles of code author and code reviewer to lie on continuous spectra, each ranging from fully human-led to fully LLM-led. Based on this human–AI interplay in code review, four possible futures of collaborative software engineering begin to take shape.