When an employee says your performance review is AI-written

You’re sitting in a review and your report says, “This reads like AI.” Your stomach drops. This guide gives you the words to steady the room, show your work, own any shortcuts, and turn the talk back to facts and a plan.

How do you open the conversation?

Don’t argue the AI point first. Center the purpose and your ownership.

Say this:

“I hear that this review felt generic and maybe AI-written. I wrote it. I own it. Let’s walk through the specific examples behind each point and make sure this reflects your work. If anything is off, I’ll fix it today.”

Then set a simple agenda: what’s working, what’s not, evidence, next steps. Keep it slow and human.

What do you actually say when they accuse you of using AI?

You need clear, calm responses and boundaries. Here are lines you can use in the moment.

• “I didn’t outsource this judgment. These are my words and my call. I’ll show you the notes and examples behind each section.”
• “I did use a company template to keep structure consistent. The assessment and examples are mine.”
• “If any phrase sounds canned, let’s replace it with plain English right now and keep the substance.”
• “We can talk about tools, but the core is your impact this cycle. Let’s check the facts together.”
• “If you feel the process wasn’t fair, I want to hear it. After we review specifics, we can talk about process changes.”

How do you show your review is grounded in specifics?

Generic language fuels suspicion. Replace it with receipts.

Do this live: open your notes. For each point, give one behavior, one context, one impact, one date.

Say:

“On March 18, in the Q2 kickoff, you presented the ops plan without risk scenarios. As a result, finance pushed back and we slipped approvals by one week. That’s why I called out planning depth. Next time, include two risk paths.”

• Tie each rating to an expectation you shared earlier. If you never shared it, own that and reset it now.
• Show work samples, Jira tickets, call notes, or emails. Keep names redacted if needed.
• When you’re wrong on a fact, say: “You’re right. I misremembered that. I’ll correct the review.” Then correct it on the spot.

What if you used AI to draft without losing trust?

If you used AI for phrasing or a checklist, don’t hide it. Be precise about what the tool did and didn’t do.

Say:

“I used a template and a writing aid to tighten formatting and avoid typos. The judgments, examples, and ratings are mine. If any wording feels off, we’ll rewrite it here so it sounds like me and reflects you.”

Know your company’s policy. If you’re unsure, say you’ll confirm and follow up. Don’t make promises about data handling you can’t back up.

• Avoid saying “I didn’t use AI” if you did for text cleanup. Trust dies on hedging.
• Keep the focus on judgment: “Tools helped with words; I’m accountable for the call.”
• If policy bans AI for reviews, stop, acknowledge it, and redo properly. Loop in HR.

How do you handle it if the review is thin or rushed?

Own it, don’t defend it. Then fix it fast.

Say:

“You’re right, parts of this are light. I rushed. That’s on me. Let’s pause ratings, walk through real examples together, and I’ll rewrite by tomorrow. We’ll then reconvene to align on the plan.”

Then actually rewrite. Thin reviews are gasoline on the AI fire.

• Add at least three concrete examples per major point.
• Swap buzzwords for actions and outcomes.
• Document the updated review and share it promptly.

How do you keep this from becoming an AI argument?

A debate about models won’t help either of you. Keep bringing it back to the work.

Use boundary lines:

• “We can spend five minutes on process questions, then I want to spend the rest on your goals and support.”
• “I won’t argue about phrasing. Let’s fix it and focus on substance.”
• “I hear the worry. After this, I’ll share how I arrived at the rating, then we’ll align on the growth plan.”

A script you can adapt

Before you walk in

• Pull concrete examples: date, context, behavior, impact, next time.
• Gather artifacts: emails, tickets, call notes, docs. Redact sensitive info.
• Map each rating to a published expectation. If missing, write it now.
• Rewrite buzzwords into plain language before the meeting.
• Decide what you’ll disclose about any templates or tools used. Check policy with HR.
• Timebox the meeting to allow review, correction, and plan setting—don’t rush it into 30 minutes. Block 60-75 minutes if possible.

Rehearse it before you say it

Run this exact exchange out loud in Rehearse so you can steady your tone, swap buzzwords for proof, and practice boundary lines before the real meeting. Rehearse gives you an AI partner who plays the other person — by voice or face-to-face — then coaches you on clarity, empathy and structure.

On your phone? Tap here →

Questions people ask