Did GitHub become more uniform after ChatGPT? Code study for 2019–2025

The author of a post on Habr tested the hypothesis that open code on GitHub has become more uniform after the appearance of ChatGPT and the widespread adoption of AI tools for development like GitHub Copilot, Cursor, Claude Code, and Windsurf — for this he analyzed platform data for 2019–2025.

Why check code for uniformity

After ChatGPT appeared and AI assistants for developers spread widely, development has noticeably accelerated: code, tests, README, comments, and small utility functions can now be obtained in seconds. But there is a less obvious side to this — if millions of developers use similar models and similar prompts, open code risks becoming more homogeneous. This is a hypothesis, not a ready conclusion: the author himself formulates the question as an open problem that needs to be tested on real data, not as a known fact in advance.

How the research is organized

The author built the analysis on GH Archive and BigQuery data for 2019–2025, studying commit messages, README contents, function names, and other textual artifacts from repositories. To compare different periods, he introduced his own metric — GitHub Uniformity Index, allowing assessment of how similar code and accompanying texts became before and after the mass adoption of AI tools in development.

The choice of the 2019–2025 time period is not accidental: it covers the years before ChatGPT appeared, the moment of its release, and subsequent years when AI assistants for code — from GitHub Copilot to Cursor, Claude Code, and Windsurf — went from niche plugins to a mass development tool. This allows searching not just for a one-time spike in metrics, but for a sustainable trend stretching over several years of technology adoption.

Why this matters for open code

If the hypothesis is confirmed, it would mean that the convenience of AI assistants comes at a price — a reduction in the diversity of approaches and styles in open repositories. For an ecosystem that has historically valued diversity of solutions to the same problem, such an effect would be a significant signal of the side effects of mass adoption of generative models in everyday development.

• Analysis period — GitHub data for 2019–2025
• Data sources — GH Archive and BigQuery
• Commit messages, README and function names are studied
• Custom metric introduced — GitHub Uniformity Index
• AI tools mentioned — GitHub Copilot, Cursor, Claude Code, Windsurf

What exactly can be measured from code on GitHub

GH Archive records public events on the platform — commits, pull requests, issues — and the connection with BigQuery allows running SQL queries across this array over many years without downloading entire repositories. This approach makes it possible to compare commit style, README structure, and function naming in slices "before ChatGPT" and "after ChatGPT" on statistically significant data volumes, not on a sample from just a few projects.

What this means

The question of code uniformity is not about aesthetics, but about whether open code is losing diversity of approaches and solutions along with the growth in development speed. The GitHub Uniformity Index methodology provides a tool to measure this effect on real data, not just assume it by eye. The way the question itself is framed shows that the discussion about AI tools for development is gradually shifting from simple "it became faster" to more subtle side effects of mass adoption.