Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

"TODO: Fix the Mess Gemini Created": Towards Understanding GenAI-Induced Self-Admitted Technical Debt

Abdullah Al Mujahid, Mia Mohammad Imran

TL;DR

The paper investigates how generative AI involvement in software development shapes self-admitted technical debt (SATD) by analyzing 6,540 LLM-referenced code comments and identifying 81 instances that acknowledge debt. It introduces GenAI-Induced Self-admitted Technical debt (GIST) as a lens to describe debt arising when developers integrate AI-generated code with limited understanding and deferred verification. The study reveals a shift in SATD composition—design debt becomes less dominant while requirement and test debts rise—and identifies AI roles in comments as Source, Catalyst, Mitigator, or Neutral, indicating nuanced human-AI collaboration and accountability dynamics. It advocates integrating AI practices into the software development lifecycle, emphasizes explainability and governance to manage AI-related debt, and calls for broader, cross-language longitudinal research to validate and extend these findings.

Abstract

As large language models (LLMs) such as ChatGPT, Copilot, Claude, and Gemini become integrated into software development workflows, developers increasingly leave traces of AI involvement in their code comments. Among these, some comments explicitly acknowledge both the use of generative AI and the presence of technical shortcomings. Analyzing 6,540 LLM-referencing code comments from public Python and JavaScript-based GitHub repositories (November 2022-July 2025), we identified 81 that also self-admit technical debt(SATD). Developers most often describe postponed testing, incomplete adaptation, and limited understanding of AI-generated code, suggesting that AI assistance affects both when and why technical debt emerges. We term GenAI-Induced Self-admitted Technical debt (GIST) as a proposed conceptual lens to describe recurring cases where developers incorporate AI-generated code while explicitly expressing uncertainty about its behavior or correctness.

"TODO: Fix the Mess Gemini Created": Towards Understanding GenAI-Induced Self-Admitted Technical Debt

TL;DR

The paper investigates how generative AI involvement in software development shapes self-admitted technical debt (SATD) by analyzing 6,540 LLM-referenced code comments and identifying 81 instances that acknowledge debt. It introduces GenAI-Induced Self-admitted Technical debt (GIST) as a lens to describe debt arising when developers integrate AI-generated code with limited understanding and deferred verification. The study reveals a shift in SATD composition—design debt becomes less dominant while requirement and test debts rise—and identifies AI roles in comments as Source, Catalyst, Mitigator, or Neutral, indicating nuanced human-AI collaboration and accountability dynamics. It advocates integrating AI practices into the software development lifecycle, emphasizes explainability and governance to manage AI-related debt, and calls for broader, cross-language longitudinal research to validate and extend these findings.

Abstract

As large language models (LLMs) such as ChatGPT, Copilot, Claude, and Gemini become integrated into software development workflows, developers increasingly leave traces of AI involvement in their code comments. Among these, some comments explicitly acknowledge both the use of generative AI and the presence of technical shortcomings. Analyzing 6,540 LLM-referencing code comments from public Python and JavaScript-based GitHub repositories (November 2022-July 2025), we identified 81 that also self-admit technical debt(SATD). Developers most often describe postponed testing, incomplete adaptation, and limited understanding of AI-generated code, suggesting that AI assistance affects both when and why technical debt emerges. We term GenAI-Induced Self-admitted Technical debt (GIST) as a proposed conceptual lens to describe recurring cases where developers incorporate AI-generated code while explicitly expressing uncertainty about its behavior or correctness.
Paper Structure (24 sections, 2 figures, 1 table)

This paper contains 24 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Data Collection
  5. Collection of LLM Referenced Comments
  6. Debt Acknowledgment Detection
  7. Resulting Dataset
  8. SATD Types
  9. Taxonomy of SATD
  10. Annotation
  11. Identifying the Role of AI
  12. RQ1: What types of self-admitted technical debt emerge in developer comments that acknowledge both LLM involvement and technical debt?
  13. Results
  14. Discussion.
  15. RQ2: How do developers attribute the role of AI in generative-AI induced SATDs?
  16. ...and 9 more sections

Figures (2)

  • Figure 1: Example self-admitted comment illustrating requirement debt
  • Figure 2: GenAI roles across SATD types