Executive summary
In the October 2026 sample of 1,736 repositories, 15.21% were classified as built with AI and 4.03% as using AI. The margin of error is 1.69 percentage points. Combined, 17.80% fell into either group.
- Built with AI: 264 repositories in the sample
- Using AI: 70 repositories
- Either group: 309 repositories, of which 25 were in both
The built-with-AI share is the highest in the seven months of history available, but the 2.47-point rise from September is only about 1.5 times the margin of error, so one month alone does not establish a rise. The using-AI share has stayed between 4.03% and 5.12% over the same period. This is an interpretation: growth appears concentrated in repositories built with AI assistance, not in repositories that add AI features.
The sample is drawn from a public dataset (GH Archive) whose coverage has degraded since 2025, and the margin of error above does not include that effect. A separate sample of 700 repositories chosen at random without GH Archive gave 14.4% built with AI (margin 2.6 points) and 4.3% using AI, consistent with this survey. That check covers October 2026 only.
Key movements
Among the repositories this site tracks, 43,807 were AI-related this month and 14,524 carried an AI coding tool signal. These counts describe our own corpus, not GitHub as a whole. The tool signals were dominated by two instruction files.
- CLAUDE.md: 7,397 repositories
- AGENTS.md: 5,983 repositories
- Cursor rules: 530 repositories
- Copilot instructions: 367 repositories
- Cline rules: 43, Windsurf rules: 38, Aider config: 11
The tracked corpus also holds 6,376 solo-developer candidates and 7,032 candidates with a public web launch. The continuity rate, meaning repositories created three months earlier that were still being pushed to, was 19.66%. It carries medium confidence.
The most-starred AI-related repository was nexu-io/open-design (TypeScript) with 103,302 stars. Next were lexmount/moli (Rust) with 19,443 and experientiallabs/experiential (Python) with 13,515. Star counts are cumulative, so they show accumulated attention, not activity this month.
Twelve-month trend
The available history covers seven months, April to October 2026, not twelve. The built-with-AI share rose from 7.51% in April to 15.21% in October, a gain of 7.70 percentage points.
- April: 7.51% built, 4.58% using
- June: 8.86% built, 4.40% using
- August: 13.04% built, 4.35% using
- September: 12.74% built, 4.42% using
- October: 15.21% built, 4.03% using
The path was not perfectly smooth. September came in slightly below August, and its sample was the smallest at 1,515 repositories. The following rise to 15.21% is larger than the stated margin of error. The using-AI share moved within a narrow band with no clear direction, and its October reading is the lowest of the seven months.
Outlook
Each monthly figure comes from a sample of roughly 1,500 to 1,800 repositories, so small changes should not be over-read. A shift of one or two points can fall within sampling noise.
- Watch whether the built-with-AI share holds above 15% in the next sample
- Watch whether the using-AI share stays in its 4% to 5% band
- Watch whether the AGENTS.md count keeps pace with the CLAUDE.md count, as a sign of convention changes
The tool-signal counts are a partial view because they only capture repositories with a visible configuration file. Interpretation of the trend should stay tentative until more months of data are available.