Canonical List of Data-Backed AI Developer Productivity and Code Quality Research from 2025-2026
GitClear's first-party research has been cited by numerous reputable sources, including
MIT Technology Review, TechCrunch and The New Stack. But we don't have the monopoly on great, data-backed research.
This page collects widely-cited third-party research on AI developer productivity and code quality between 2024 and
2026. If you have a link to AI research that has changed your thinking, let us know?
|
Research Title & Source |
Publication Date |
Key Takeaways |
 |
GitClear Research
|
January 2026 |
- Heavy AI users are generating 4-10x more durable code than non-AI users
- Those same users also generated 9x more code churn than non-AI users
- The 4x jump in productivity reflects truths about who is using AI most, as of early 2026
|
 |
GitClear Research
|
2025 |
- Analyzed 211 million lines of code from 2021 to 2025
- Finds that percent of moved/refactored code plummeted from 25% in 2021 to less than 10% in 2025
- Over the same period, copy/pasted (duplicated) code rose from 8% of changes to 18%
|
 |
Google Research
|
2025 |
- Most “lead time” is still waiting, not building: ~21% flow efficiency. In the value-stream example: ~24.5 hours active work vs ~92.5 hours waiting (total ~117 hours lead time) → ~21% flow efficiency.
- Estimated impact of AI on "Software Delivery Instability" is 0.1x, second only to the increase in "Individual Effectiveness" (at 0.17x)
- Using AI to resolve a microservice can take less than 5 minutes of process time, but still incur a 5 day cycle time due to review waits
|
 |
Stack Overflow
|
2025 |
- Daily AI use is common: 47.1% use AI tools daily (plus 17.7% weekly and 13.7% monthly/infrequently)
- Professional developers are even more “daily”: 51% of professional developers use AI tools daily.
- Trust is the constraint: only 3.1% “highly trust” AI accuracy; overall 33% trust vs 46% distrust AI outputs.
|
 |
GitClear Research
|
2025 |
- Analyzes 70,000 developer-years of data to quantify the impact of AI tools on developer productivity
- Median developer productivity has increased 9% from 2022 to 2025
- Among developers averaging 500+ annual commits, the median productivity increase vs 2022 is 14.1%
|
 |
JetBrains
|
2025 |
- 68% of developer anticipate that AI proficiency will become a job requirement
- The most used AI tool in 2025 is still ChatGPT web/desktop/mobile apps (41%) - ahead of any IDE or plugins
- Most likely to delegate to AI (in descending order): writing boilerplate, searching the internet, converting code to other languages, writing documentation, summarizing recent changes
|
This page will be updated monthly throughout 2026 with new research as it becomes available. Bookmark and check back later for more?
