Methodology

What is counted

For each tool with a verified GitHub bot, we ask the GitHub REST Search API how many public pull requests created in one ISO week were formally reviewed by that bot. The exact query is:

is:pr is:public created:<Monday>..<Sunday> reviewed-by:<bot login>

A secondary column uses the same query with commenter: instead of reviewed-by:, which also counts PRs where the bot only left a comment (for example a summary). Each request is sent to https://api.github.com/search/issues with per_page=1; we record GitHub's total_count, incomplete_results, the HTTP status and the request time. Every number on this site links to its query.

Cohort and schedule

• We measure the ISO week that ended at least 7 days before the run, so reviews that arrive a few days late are included.
• The refresh runs weekly (Tuesday 06:00 UTC). Monthly reports cover a calendar month and are measured after the 8th of the next month (from 00:00 UTC on the 9th); an earlier measurement is never published.
• Requests are unauthenticated, sent one at a time at least 15 seconds apart, below GitHub's published limit of 10 searches per minute. We never scrape GitHub's HTML pages.
• If GitHub answers 403 or 429, the run stops, is saved as partial, and the site keeps showing the last complete week with a stale badge. A failed request is stored as empty, never as 0, and a total GitHub flags as incomplete_results is shown as unavailable, not as a number.
• Each complete weekly snapshot is immutable and kept in the public data files.

Latest complete week: 2026-W39, measured 2026-10-07.

What the numbers are not

• Not unique users or customers. A count is pull requests, and one customer can open many.
• Not private repositories. Only public PRs are visible; vendors with mostly private usage look smaller.
• Not quality. A review being posted says nothing about whether it was useful. See independent benchmarks for that.
• Not market share. One PR can be reviewed by several bots, so counts cannot be summed or turned into shares.
• Not exhaustive. GitHub search totals are what GitHub reports for the query at that moment and can drift slightly.

Bot identity checks

A tool gets a number only when its bot passes all of these:

1. The login resolves through GET /users/<login> to an account of type Bot; we store its numeric ID.
2. A first-party source ties that bot to the product: the vendor's docs or install link, or the GitHub App page published by the vendor's own organisation.
3. An audit samples 3 public PRs from the reviewed-by: results and confirms via GET /repos/…/pulls/…/reviews that each contains a review by that numeric ID. Only the pass count is stored.
4. Every month the numeric ID and type are re-checked; a mismatch downgrades the tool to unverified.

Tools that post as github-actions[bot] or through the user's own token are not attributable (12 tools). Bot accounts that serve several products are marked shared and are not ranked (10). Tools whose bot has not passed yet are unverified (8). None of these is shown as 0.

Facts: pricing, platforms, self-hosting, licence

Every fact comes from the vendor's own website, docs, repository or official marketplace listing, with the source link and the date it was checked. We do not copy prices from review sites. Facts older than 30 days are re-checked by hand, one page at a time. When a fact is not published, we say so instead of guessing.

A tool is listed if its main function includes AI review of pull or merge requests and it is generally available. Listing is free and is not for sale.

Integrity rules for sponsorship

• Payment never changes the ranking, the counts, the verification status, or which facts are shown.
• Sponsored cards carry a “Sponsored” label and sit outside the ranked table.
• No paid tier is called “verified”. Verification is free and applies to every tool the same way.

Details: advertise.

Data and privacy

We store only aggregate counts and bot identities. PR titles, authors, repository names and any other personal data are never stored; an automated test fails the build if a data file contains such fields. GitHub data is used under GitHub's API terms for aggregate public research; this site does not resell API access or data.