Sonar's 2027 State of Code Developer Survey, published on 12 November 2026, finds that developers say AI now generates or significantly assists 65% of the code they commit, up from 42% a year earlier. The survey of 1,026 developers and engineering managers also found that 76% generate AI code faster than they can verify it.
What the survey found
Sonar asked the same population a year ago how much code AI would write. Respondents then expected 55% within a year and 65% within two, and they reached 65% in one. The share of teams that use AI coding tools for business-critical or mission-critical services rose from 58% to 76%.
Checking has gone the other way. The share of respondents who always check AI code before committing it fell from 48% to 39%, and only 23% say their team has well-defined, AI-specific guidelines or automated checks for AI-generated code. Only 36% say their team measures AI's productivity impact and can see the results.
Where teams are heading next
Sonar reports that 5% of respondents already create or use agentic software factories, meaning systems of agents that work through development tasks with little human input, and 34% expect to within a year. Of the respondents, 77% say strong verification tools raise their trust in AI code, and teams that give coding agents clear standards and controls are more likely to report an ROI of 50% or more.
Sonar sells code verification tools and the results are self-reported, so the numbers describe what developers believe about their own work. The direction of the findings matches other recent studies on review load, which show code volume rising faster than the capacity to read it.
What it means for engineering teams
If code is produced faster than it is checked, the limit on delivery moves to review, testing and the people who can judge whether a change is correct. Teams that write down the standards for agents, run automated checks before merge and measure results are in a stronger position than teams that rely on a final human read of every pull request. That review work depends on senior engineers who know the codebase well enough to spot a plausible but wrong answer.
What it means when you hire engineers
With AI involved in nearly two thirds of committed code, the useful hiring question is how a candidate verifies agent output: what they accept, what they send back and how they catch a defect that passes the tests. Every engineer we place is tested by an AI-native tech lead who works with AI every day, and you can read how that works on how we vet. If you are adding capacity for review-heavy work, you can read about our AI-native engineers.
Sources
- Sonar, Sonar Survey: AI Makes Code Abundant, but Code Confidence Remains Scarce, 12 November 2026
- Sonar, AI hit 65% of committed code a year early, 12 November 2026
- Stackchief, AI Code Verification Survey 2026, 14 November 2026




