Our AI code review data after 200 pull requests says two things. Precision matters more than coverage. And the most useful learning happens after the code is merged.

Revix AI has now reviewed 200 pull requests. We tracked every comment and every merge from day one.

AI code review results after 200 PRs

Revix AI caught 78 issues before merge. Of 216 review comments posted, 151 flagged real issues the team acted on. That is a 70% signal rate.

issues caught before merge: 78
comments that flagged real issues: 151/216
signal rate across all review comments: 70%

These were real issues, not nitpicks. A query that would have missed an index. A retry loop without backoff. An endpoint that skipped the permission check every other endpoint in the module ran.

Signal rate is the metric we watch most, because noise is how AI code review fails.

Precision beats coverage

A noisy bot gets ignored. It comments on everything: style opinions, restated diffs, suggestions that contradict how the repo works. After a week, developers scroll past. After a month, someone asks to turn it off.

A reviewer that speaks up 216 times and is right 151 times builds a better habit: people read the comments.

So we tuned Revix AI to stay quiet unless it has something worth saying.

The shape of a comment matters as much as the volume. Each comment opens with a Conventional Comments label such as issue: or suggestion:, then names the specific behavior and a way out. Our guide to writing a great PR review comment covers the format.

Merged code teaches the reviewer

After a pull request merges, Revix AI reads the diff again and asks one question:

Did this change what "correct" looks like for this repository?

Sometimes the answer is yes. The team adopted a new error-handling pattern. A module was restructured. A convention that held for a year stopped applying. Merged code shows what the team actually accepts, so it is the best learning signal available.

Across 129 completed merge analyses, that feedback loop has:

new rules created from merged code: 87
existing rules refined: 47
rules retired as the codebase evolved: 3

The Revix AI rules dashboard listing derived rules grouped by impact and by app — Every merge analysis can create, refine, or retire rules, so the reviewer's standard tracks the codebase.

Retired rules matter too

The 3 retired rules matter as much as the 87 new ones. They no longer matched how the codebase worked. A static reviewer would keep enforcing them, flagging correct code as wrong. Each false positive costs a little of the team's trust.

Most AI code review tools learn once

The typical tool scans your repository once during onboarding, maps your conventions and starts reviewing. That map is accurate on day one. Then the repository changes, and the reviewer's view stays frozen at setup.

The result is slow drift. Six months in, the reviewer enforces a standard the team dropped two quarters ago. False positives pile up, and the tool ends up ignored like the noisy bots.

This is a second version of the visibility gap in repo-aware vs. diff-based review. There, the reviewer can't see past the diff. Here, it can't see past onboarding day.

Ask any vendor how their reviewer keeps up with a changing codebase. We compare how CodeRabbit gets its standards in CodeRabbit alternatives compared.

Takeaways from 200 reviews

200 reviews is early, but the data points one way:

High-precision comments get read. Noisy ones train people to scroll past.
Post-merge analysis is the strongest learning signal, because merged code is what the team accepts.
Rules need retiring, not just creating. A ruleset that never prunes becomes its own source of false positives.

We think AI code review is heading toward reviewers that keep learning from every merge.

New to the topic? Start with what AI code review is and how it works. The launch post covers what Revix AI does: PR summaries, Conventional Comments labels and ticket-aware reviews.

To see it learn from your next merge, connect a repository at revix.ai. Every team starts with a 14 day free trial, no card required.