What Is AI Code Review and How Does It Work?

What AI code review is

AI code review is an automated pass in which a language model comments on a pull request the way a human reviewer would. It reads the change, plus whatever part of the repository the tool can see. Then it flags bugs, convention breaks, and risks before a person opens the diff.

How much of the repository it can see is the key variable. It decides what the tool can find, so compare it first when you evaluate one.

How AI code review works

Every useful tool runs four stages.

1. Trigger. A webhook fires when a pull request opens or gets new commits. Nothing runs on a schedule.
2. Context. The tool pulls the diff. Some products also retrieve related files: the modules the change imports, conventions derived at onboarding, sometimes the linked ticket.
3. Reasoning. A model reads the change against that context and produces candidate findings.
4. Filtering and posting. The tool drops weak findings and posts the rest as inline comments plus a summary on the pull request.

Products differ most at stage 2. Stage 4 decides whether your team keeps paying attention. A tool that skips filtering posts everything it thought of, and that is how review bots end up muted.

What it catches

It catches mechanical defects, applied the same way to every diff. Examples: a missing null guard, a retry loop with no backoff, a swallowed error. Or an endpoint that skips the permission check every other endpoint in the module performs.

The strongest category is convention drift. That is code that compiles and passes tests but solves a problem differently from the rest of the repo.

This kind of finding is why AI review runs alongside linters instead of replacing them. A linter enforces rules someone wrote down. An AI reviewer also works from conventions nobody documented, which is most of them.

What it misses

It does not decide whether the feature should exist. It can't judge whether a deadline justifies a shortcut, or whether the team can maintain an abstraction after its author leaves.

That context lives in people, not in the repository, so a bigger model does not help. We covered that boundary in Is AI Code Review the New Standard, or Just Hype?

Diff-based vs repo-aware review

A diff-based reviewer reads only the files a pull request touches. It is fast, cheap, and needs no setup. Anything outside the diff does not exist to it.

A repo-aware reviewer reads the change against shared helpers, internal frameworks, and decisions made elsewhere. It can ask whether the change fits the system, not only whether it is valid alone.

The gap shows up in specific places:

• An abstraction reimplemented three services over.
• A new endpoint that bypasses the shared authorization middleware.
• A query that skips the index every comparable query uses.

Worked examples are in Repo-Aware AI Code Review vs. Diff-Based Review.

How to evaluate a tool

Feature tables don't help here, because every vendor claims the same things. Ask three questions instead:

1. What is your signal rate? Of the comments it posts, what share does a team act on? A vendor without this number has not measured it. Ours is 70% across 216 comments on 200 pull requests.
2. How much context do I get on my plan? Vendors often gate repository awareness behind the top tier. The demo and the plan you buy can behave differently.
3. What happens after merge? A tool that scans once at onboarding reviews against your repo as it was on setup day. Ask how its rules update.

Then run it on a pull request you already reviewed. Did it find what you found, and did it invent anything? We ran that comparison in CodeRabbit alternatives compared. To see nine tools side by side, compare AI code review tools.

Revix is a repo-aware AI code reviewer. It extracts patterns from your codebase and turns them into review rules. The rules update as the repo changes, and every PR is reviewed against them.

See how Revix does it on our AI code review tool page. Every team starts with a 14 day free trial, no card required.