These 14 performance review prompts help you write a self-evaluation, a manager review, peer and upward feedback, and next-period goals from your real evidence, not invented praise. Every prompt asks the model to mark gaps instead of making things up, and the last sections cover a bias check and what not to paste into an AI tool.
How to use these prompts
Each prompt works in ChatGPT, Claude, Gemini or Microsoft Copilot. Replace the blanks in double braces, or keep the default after the colon: {{period:the last 12 months}} means the last 12 months unless you change it.
The prompts are written around one rule: the model only uses facts you give it. Where a prompt says {{notes}}, paste your raw material: a brag document, project list, shipped work, metrics, feedback you received, or your company's competency framework. Where the evidence is thin, the prompt tells the model to write [NEEDS EVIDENCE] so you can fill it in. That marker is the most useful part. A review with real numbers beats a polished one with none.
Before you paste anything, read the privacy section near the end of this page. Your company may have rules about which AI tools you can use for HR material.
How do you use AI for performance reviews without making them generic?
Generic reviews come from generic input. If you ask "write my self-review as a software engineer", the model has nothing to go on, so it writes what every review says. Three habits fix most of this:
- Start from evidence. Paste notes with dates, numbers, names of projects and outcomes. Microsoft's guide to writing Copilot prompts calls this the source: the information the assistant should use.
- Ask for a fixed structure. Your company's form, a competency list, or "impact, how, what's next" all work. Structure stops rambling.
- Review every line. Microsoft's own advice for Copilot is "Review and verify responses you get from Copilot." That goes double for anything that affects pay or promotion.
Use the model to organize, tighten and check. Keep the judgment, the rating and the final wording yours.
Which prompts help write a self-evaluation?
1. Turn your notes into an accomplishments list
Start here. Most people forget half of what they did, and this pulls it out of messy notes.
Below are my raw notes from {{period:the last 12 months}} as a {{role:product designer}}. They include projects, messages, metrics and feedback, in no particular order.
Turn them into a list of accomplishments. For each one give: what I did, the result, and the evidence (a number, a date, a quote or a link from my notes). Group them by {{grouping:project}}. Only use facts from my notes. If an item has no clear result or evidence, keep it but add [NEEDS EVIDENCE]. Do not add accomplishments that aren't in the notes.
Notes:
{{notes}}
2. Draft the self-evaluation against your company's form
Use your company's real questions or competencies so the draft fits the form you have to fill in.
Draft my self-evaluation using my accomplishments list and my company's review questions below.
Answer each question in {{length:120 to 180}} words, in the first person, in a {{tone:confident but factual}} tone. Lead each answer with the most important result. Use specific examples from the list, with numbers where I have them. Where I have no evidence for a question, write [NEEDS EVIDENCE: what kind of example would help] instead of filling the space.
Review questions:
{{questions}}
Accomplishments:
{{accomplishments}}
3. Find the missing numbers
For drafts full of "improved" and "helped" with no scale. The model asks questions; you supply the answers.
Read my self-review draft below. Find every claim that would be stronger with a number, a before and after, a time saved, or a named outcome.
For each one, quote the sentence and ask me one specific question that would let me add the missing fact (for example, "How many support tickets did this reduce per week?"). Don't suggest numbers yourself. List at most {{max:10}} questions, most important first.
Draft:
{{draft}}
4. Write growth areas honestly
The section people find hardest. This keeps it honest without sounding like a confession.
Help me write the "areas for growth" part of my self-review. Here is what I think I struggled with this period, and any feedback I received:
{{struggles}}
Pick the {{count:two}} most important areas. For each, write two or three sentences: what happened, what I learned, and what I will do differently next period. Keep it specific and calm. Don't turn weaknesses into disguised strengths, and don't add problems I didn't mention.
Which prompts help managers write reviews?
If you manage people, check what your company allows before you start. Use the tool and account your employer approves, and remove details you don't need, like full names. The prompts below use "Employee A" for that reason.
5. Draft a review from evidence
For the first draft of a direct report's review, built from your notes and their self-review.
I'm writing a performance review for Employee A, a {{level:mid-level}} {{role:software engineer}} on my team, for {{period:the last 12 months}}.
Below are my notes, their self-review and feedback from peers. Draft the review in these sections: summary, key contributions, strengths, areas for growth, and focus for next period. Base every statement on the material below and reference the example it comes from. Where my notes and their self-review disagree, point it out instead of picking a side. Mark any section where the evidence is thin with [NEEDS EVIDENCE]. Do not suggest a rating.
My notes:
{{manager_notes}}
Self-review:
{{self_review}}
Peer feedback:
{{peer_feedback}}
6. Check a draft against your rating guide
For calibration prep. The model compares your words to your company's definitions; you decide the rating.
Below is my company's rating guide and my draft review for Employee A. I'm considering a rating of {{rating:Meets expectations}}.
For each level in the guide, say which parts of my draft support it and which don't. Point out where my written evidence seems stronger or weaker than the rating I'm considering. Don't tell me what rating to give; help me see whether my evidence matches the words I'm using.
Rating guide:
{{rating_guide}}
Draft review:
{{draft}}
7. Phrase difficult feedback
OpenAI's ChatGPT for managers page includes a similar use case: asking for a few ways to phrase constructive feedback, with the pros and cons of each.
I need to give Employee A constructive feedback about {{issue:missing deadlines on the last three projects}}. Here are the specific examples: {{examples}}.
Give me {{options:three}} ways to say it in the review conversation, from most direct to most gentle. For each, write the exact words I could use, then list one pro and one con. Every option must name the specific behavior and its impact, and end with a clear expectation for next period. Avoid labels about personality.
8. Plan the review conversation
For the meeting itself, which matters more than the document.
Help me plan a {{length:45}} minute review conversation with Employee A. Their review summary is below.
Give me an agenda with times, two open questions to start, the key message I must get across, how to respond if they disagree with {{sensitive_point:the growth area on communication}}, and how to close with agreed next steps. Keep the tone supportive and direct.
Review summary:
{{summary}}
Which prompts write peer and upward feedback?
9. Peer feedback that's specific
For 360 forms where you want to be useful, not just nice. It uses a situation, behavior, impact structure.
Help me write peer feedback for a colleague I worked with on {{project:the billing migration}}. Here are my rough thoughts:
{{thoughts}}
Write {{count:two}} strengths and {{count_growth:one}} suggestion. For each, use this order: the situation, what they did, and the impact it had. Keep each to three sentences or fewer. Use only the examples in my notes. If one of my points has no example, ask me for one instead of writing it.
10. Upward feedback for your manager
For giving your manager honest feedback that they can act on.
I want to give my manager upward feedback. Here is what works well and what I'd like to change:
{{thoughts}}
Write it in a respectful, direct tone. Start with what I value, with a concrete example. Then give {{count:one or two}} requests, each framed as a specific change I'd like to see and why it would help the team. No sarcasm, no hedging, and no claims beyond what I wrote. Keep it under {{length:200}} words.
11. Answer one review form question
For the short text boxes on review forms, where you have 500 characters to make a point.
Answer this review form question in no more than {{limit:500}} characters: "{{question}}"
Use these facts only: {{facts}}
Lead with the most important point, include one specific example, and cut any filler. Give me two versions, one more formal and one more conversational.
Which prompts set goals for the next period?
12. Goals from your review
For turning growth areas and team priorities into goals someone can actually check at the next review.
Based on my review summary and my team's priorities for {{next_period:next year}}, suggest {{count:three to five}} goals for me.
For each goal give: the goal in one sentence, why it matters to the team, how we'll know it's done (a measure or a clear deliverable), and a target date. Mix outcome goals with at least one development goal. Flag any goal that depends on someone else.
Review summary:
{{review_summary}}
Team priorities:
{{priorities}}
13. Development plan
For a growth area that needs more than one goal, like moving toward a promotion.
I want to grow toward {{target:a senior engineer role}}. My company's expectations for that level are below, along with my current strengths and gaps.
Build a {{months:six}} month development plan: three focus areas, two concrete actions for each (projects, stretch work, learning, mentoring), and the evidence I should collect along the way so my next review is easy to write. Keep it realistic for someone with a full workload.
Level expectations:
{{expectations}}
My strengths and gaps:
{{strengths_and_gaps}}
If your team works in objectives and key results, the OKR prompts help turn these into team-level OKRs.
How do you check a review for bias and vague language?
14. Bias and vague-language check
Run this on any review before you submit it, whether you or the AI wrote it.
Review the performance review below for problems. Check for:
1. Vague phrases with no example behind them (like "great attitude" or "needs to be more strategic").
2. Comments about personality instead of behavior.
3. Words that can read as biased or coded, such as "abrasive", "emotional" or "aggressive".
4. Claims about results with no evidence.
5. Recent events weighted more than the rest of the period.
Quote each problem, explain why it's a problem in one line, and suggest a specific rewrite or the question I need to answer to fix it. Don't change the overall assessment.
Review:
{{review}}
What should you never paste into an AI tool during review season?
Reviews are full of personal data about you and other people. Before you paste anything, check your company's AI policy and use the tool it approves. Then follow these rules:
- Don't paste confidential employee data into a personal or consumer account when your company's policy says not to. That includes salary, ratings, health or leave details, disciplinary records and anything about colleagues who didn't agree to it.
- Remove what the model doesn't need. Use "Employee A" or initials, cut employee IDs, and summarize sensitive situations instead of pasting raw messages.
- Know the difference between work and personal accounts. For example, Microsoft's Copilot Chat privacy page says work prompts are covered by enterprise data protection and "Prompts and responses aren't used to train the underlying foundation models." Personal accounts of any assistant have their own settings, so read them before using one for work.
- Use chats that aren't remembered when that's allowed. In Claude, incognito chats are not saved to memory: "Claude won't remember your chats, so they won't be saved to Claude's memory or your chat history." That limits what carries over, but it doesn't replace your company's rules.
- Keep the decision human. AI can help you write and check a review. It shouldn't decide a rating, a promotion or a pay change.
Keep your review prompts for next time
Review season comes back every year, and so does the scramble to find last year's prompts. Kipt keeps prompts as Markdown files in a private GitHub repo you own, with {{name:default}} blanks like the ones on this page, tags and Cmd+K search, so your self-review and feedback prompts are ready next cycle. Keep the evidence itself, your notes and drafts, wherever your company says it belongs.
For the emails that follow a review, like a thank-you or a follow-up on agreed goals, use the email writing prompts. For notes from the review meeting, try the meeting notes prompts.