← News

Models & toolsSources: arXiv · 13 Nov 2026

Study links GPT assistance to faster GitHub pull request resolution

Researchers report that pull requests assisted by GPT were resolved in a median of 9 hours versus 23 hours, based on 25,473 GitHub pull requests across 9,254 projects.

In plain words

A research team studied whether developers using GPT, a large language model, close their GitHub pull requests faster. Pull requests are proposals to merge code changes that teammates must review. The study found GPT-assisted proposals were resolved in a median of 9 hours compared with 23 hours for the others.

What was announced

The authors, writing on arXiv, assembled a dataset of 25,473 pull requests from 9,254 GitHub projects and identified GPT-assisted cases using a semi-automated heuristic that combined keyword detection, regular expression filtering, and manual verification, reaching 95% labeling accuracy. Statistical modeling, including multiple linear regression and a Mann-Whitney U test, compared assisted and non-assisted pull requests at the overall level and across review phases.

The reported results show GPT-assisted pull requests cut median resolution time by more than 60%, review time by 33%, and waiting time before acceptance by 87%. In a sample of 300 assisted pull requests, developers mainly used GPT for code optimization (60%), bug fixing (26%), and documentation updates (12%). The authors conclude that early adoption of GPT can substantially boost the effectiveness of the pull request process, though the findings rest on their heuristic labeling rather than direct confirmation of every assisted case.

What it does not show

  • The evidence is the paper's abstract only, so the full methodology, dataset construction, and robustness checks cannot be verified from this excerpt.
  • The 60%, 33%, and 87% figures are the authors' own statistical findings and have not been independently confirmed.
  • GPT-assisted cases were identified by a semi-automated heuristic with 95% labeling accuracy, so some pull requests may be misclassified.
  • The usage breakdown comes from a sample of 300 assisted pull requests, a small subset of the full dataset.

For security and IT teams

Scope
The study covers GitHub pull request workflows and phase-specific review stages, not general software development tasks.
Status
The paper is a v2 cross-listed arXiv abstract dated November 2026 and has not been shown to be peer reviewed.

Sources

Source-support review completed 14 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.

Coverage: arXiv: abstract.

Updates and corrections

No corrections so far.

First published 14 Nov 2026

New to the subject? AI from the Ground Up starts from the beginning.