Catch recency bias, coded language, and unfair ratings before you hit send
Paste a draft or connect your tools, and PerfCopilot flags the six bias patterns that skew reviews — recency, halo, gendered language, and more — each tied to the evidence that should anchor the rating instead.
Check my review free — up to 5 seats → · no credit card
✓ Scans your actual draft · ✓ Every flag tied to evidence · ✓ Works inside Lattice / 15Five
The 6 bias patterns PerfCopilot flags
Most pages that rank for this explain these biases. PerfCopilot finds them in your draft and shows you the fix:
| Pattern | What it looks like | What PerfCopilot does |
|---|---|---|
| Recency bias | The review leans on the last 3 weeks | Surfaces evidence from the whole period so early wins count |
| Halo / horns | One trait colors the whole rating | Flags ratings unsupported by cited examples |
| Gendered / coded language | "abrasive," "bossy," "emotional" | Highlights the phrase + suggests a neutral rewrite |
| Vague filler | "great team player," "works hard" | Flags unfalsifiable praise with no evidence behind it |
| Similarity bias | Higher marks for people like the manager | Catches uneven language across the team |
| Central tendency / leniency | Everyone gets a 3, or everyone gets a 5 | Shows rating spread vs. the team for calibration |
The trust gap this closes: 61% of employees felt their last performance review was unfair — most of it down to hidden bias. A checker that runs before you share is how you fix it at the source.
The flags aren't guesses. Each one is tied to real activity PerfCopilot already pulls in to draft the review from evidence — merged pull requests from the GitHub integration, shipped tickets, decisions — so when the checker says a rating isn't supported, it can show you exactly what is and isn't there.
Explainers vs. an actual checker
| The "types of bias" articles | PerfCopilot | |
|---|---|---|
| Lists the bias types | ✓ | ✓ |
| Scans your actual draft and flags them inline | ✗ | ✓ |
| Ties each flag to real GitHub / Jira / Slack evidence | ✗ | ✓ |
| Calibrates ratings across the whole team | ✗ | ✓ |
| Runs inside your existing HR platform | ✗ | ✓ |
| What you leave with | a reading list | a fixed, defensible review |
Why this beats a calibration meeting alone
Calibration catches bias after every manager has already written biased drafts. PerfCopilot catches it while you write — pattern by pattern, tied to evidence — so the draft that reaches calibration is already fair and already defensible. See how it works or check pricing.
Check my review free — up to 5 seats → · no credit card
FAQ
Is the performance review bias checker free? Yes — free for teams up to 5 seats, no credit card. Pro is $4.99 per seat per month billed annually and adds unlimited seats, all 18 integrations, and SSO.
What biases does it check for? Recency, halo/horns, gendered or coded language, vague filler, similarity bias, and central-tendency/leniency — flagged inline in your draft with a suggested fix.
How is this different from an article about bias? The articles explain the bias types. PerfCopilot scans your actual review, flags each pattern where it appears, and ties the rating to real evidence so it is defensible, not just informed.
Does it see private messages? No. PerfCopilot reads activity metadata read-only and cites artifacts such as PRs, tickets, and threads; message and call content is used only with explicit opt-in.
Do I have to switch off Lattice or 15Five? No. It checks and drafts the review, then exports into the HR platform you already run.