65% of committed code is now AI-assisted. Only 39% say it's always checked.

What the survey says. Sonar surveyed 1,026 developers and engineering managers worldwide. Per its press release and findings page:

• AI now generates or significantly assists an average of 65% of the code respondents commit. A year ago, respondents put it at 42% and expected 65% within two years.
• 76% say they generate AI code faster than they can verify it.
• 53% now spend more time reviewing code than creating it, up from 18% before they adopted AI tools.
• Only 39% completely agree that AI-generated code is always checked before it is committed, down from 48% last year. 94% don't fully trust it.
• Only 23% have well-defined, AI-specific guidelines or automated checks for AI code.
• 76% say their teams use AI coding tools for business-critical or mission-critical services.

The catch. This is a vendor survey. Sonar sells code verification and review tools, so a finding that "review is the bottleneck" fits its pitch. The numbers are self-reported answers, not measured commits. Read them as a signal of how teams feel, not an audit.

Our take. The bottleneck moved from writing code to checking it. If an agent can open a pull request in minutes, the check can't be a person skimming the diff at the end. Before the next agent PR merges:

• Run the linter, type check and tests in CI, and make them required.
• Add a security scan for secrets and known-vulnerable dependencies.
• Write down what "done" means for the agent (tests added, no new warnings) so the check has a target.

Fast code that nobody checked is not fast. It's debt with a delay.

Not sponsored. MODEL has not tested Sonar's products for this story. -- MODEL Editors