Greptile's main proof is dated. The benchmarks page is titled "2025" and says the run was done in July 2025; the CodeRabbit comparison is titled "(2025)" and quotes the same run; the agent leaderboard says "Updated monthly" but still shows April 2026. Since then Greptile has shipped v4, v5, TREX and the Plus and Apex tiers, which caught bugs Base missed. The new product marketer owns launches and the site copy, and a buyer comparing tools sees year-old numbers next to rivals' fresh claims.
First action
Re-run Base, Plus and Apex on the 50 existing PRs this week.
How to measure it
Trials from benchmark and comparison pages
- Now
- –
- Target
- +30%
- By
- 60 days
Steps
Freeze the protocol
Same 50 bugs, same scoring, add tiers.
Research Engineer 2 days
Run Greptile tiers
Base, Plus and Apex on all 50.
Research Engineer 3 days
Run rivals
Current hosted plans, default settings.
Research Engineer 1 week
Verify catches
Two reviewers per finding.
Research Engineer 1 week
Write the page
Results, method, PR links, date.
Founding Product Marketing Manager 1 week
Update quotes
Homepage, comparison pages, title tags.
Founding Product Marketing Manager 2 days
Automate leaderboard
Monthly snapshot job and date stamp.
Growth Engineer 1 week
Launch
Blog, Hacker News, X and the sales deck.
Founding Product Marketing Manager 1 day
Why it works
Re-run the same 50-bug protocol on current versions of all five tools, plus Greptile Plus and Apex, and publish "AI Code Review Benchmarks 2026" with the method, the raw PR links and the date. Update the CodeRabbit and Bugbot pages and the homepage claims from the new run. Automate the agent leaderboard snapshot so it really updates monthly, and put a "last run" date on every proof page.
Engineering buyers check the date before the number. A fresh, reproducible run with the new tiers is also the clearest launch story for Plus and Apex and a reason for press and Hacker News to look again.
Ready to use
ChecklistProof page refresh list
[ ] /benchmarks retitled 2026, run date shown [ ] Base, Plus and Apex as separate columns [ ] Raw PR links and scoring sheet [ ] /greptile-vs-coderabbit and /greptile-vs-bugbot updated [ ] Homepage title claims checked against the new run [ ] Agent leaderboard on a monthly job [ ] "Last updated" on every proof page
Risks and how to handle them
A rival scores higher on one category
Publish it anyway; honesty is the point of the page.
Rivals dispute settings
Publish configs and invite re-runs.
Evidence
- Your websiteBenchmarks page, titled "AI Code Review Benchmarks 2025": 5 tools (Greptile, Bugbot, Copilot, CodeRabbit, Graphite) on 50 real bugs from Sentry, Cal.com, Grafana, Keycloak and Discourse; "Greptile led with an 82% catch rate"; "this evaluation was conducted in July 2025, and these tools evolve quickly, so performance may change over time". Since then Greptile has shipped v4 (March 2026), v5 (August 2026), TREX and the Plus and Apex tiers.Source
- Your websiteComparison pages: "Greptile vs CodeRabbit: AI Code Review Tools Compared (2025)" cites the same 50-PR run ("Greptile catches >50% more bugs"); there are also Greptile vs Bugbot, vs Claude Code and Codex, "Build vs buy", and content pages on Greptile, CodeRabbit and Graphite alternatives. There is no page comparing Greptile with GitHub Copilot code review or Qodo.Source
- Your websiteAgent leaderboard: "Updated monthly", but in late October 2026 it shows the "april 2026 update". It tracks agent-written PRs (27.6% of opened PRs in April 2026, up from 0.86% in February 2025) and revert rates per 1,000 merged PRs by author (Codex 1.19, Claude 1.80, human 2.72, Cursor background agents 3.41, Devin 3.50).Source
- Your websiteBlog, "Introducing Greptile Plus and Apex" (September 25, 2026): Plus (3 credits) "A deeper review for PRs that need extra scrutiny"; Apex (10 credits) "Our most intensive review for large, complex PRs". On a celestia-core PR, Plus and Apex caught bugs that Base missed; at Greptile, Apex "caught so many bugs that we changed how we run deploys". Teams can "set rules to use either tier automatically for large PRs or high-risk code".Source
- HiringFounding Product Marketing Manager (Growth): "every launch needs sharp content that makes developers and the leaders who buy for them get it immediately"; owns "product launches end to end, from positioning and messaging to the content that ships them (video scripts, landing pages, emails, docs, etc.)", "the copy across greptile.com", thought leadership for engineering leaders, and will "Formalize Greptile's voice, tone, messaging, and positioning". The post says Greptile has "Raised $30M from Benchmark, Y Combinator, Paul Graham, and Initialized".Source