How to Write a Performance Review With AI (2026 Guide)

Most performance reviews are lost months before anyone opens the form. By October you’re trying to rebuild a year from chat threads, half-remembered 1-on-1s, and a few notes scattered across three different apps. The form is only the last step. Whether it reads as fair depends on what you wrote down while the work was happening.

That’s why learning how to write a performance review starts well before review season. Gallup found that only a small minority of employees strongly agree their reviews inspire them to improve, and that people who get feedback every week are several times more likely to call it meaningful and to say they’re motivated to do great work. A review lands better when it confirms conversations you’ve already had.

AI fits into this in a narrower way than the marketing suggests. It can sort a year of notes, group examples by goal, tighten a clumsy draft, and point out claims you can’t back up. It can’t watch your team work, weigh which contributions mattered most, or make the call on a rating. This guide covers the whole performance review process, from the notes you start in February to the follow-through after the meeting, with AI placed where it earns its keep.

The Review Is Built All Year

A review that holds up rests on dated examples, coaching conversations, progress against goals, and next steps you agreed on together. A review that falls apart rests on memory, mood, and whatever happened in the last six weeks.

The idea is old. The Civil Service Reform Act of 1978 made performance appraisal a formal part of federal management and required that employees be judged against job-related standards they knew at the start of the period (U.S. Office of Personnel Management). Joint guidance from SHRM and SIOP makes a similar case for a small number of goals that clearly connect to what the organization is trying to do. Both point to the same habit. Judge the work against expectations you set in advance, and leave personality, popularity, and moving standards out of it.

Capture it while it happens

Good review writing depends on one habit. When something happens that you’d want to mention in a review, write it down that week. A running doc works fine. So does your task tracker or the notes app you already open every day. What matters is that you can find it again in eleven months.

Keep each note simple, with the date, the situation, what the person did, and what changed because of it. “Handled the customer escalation” won’t help you much in November. “On March 12, rerouted the priority ticket, pulled support and engineering into one thread, and had the account working again before the client’s renewal call” gives the review something solid to stand on.

Remote and hybrid teams need this more than anyone. Time zones, async work, and coordination nobody sees all make memory a weak source. A written trail keeps the review from tilting toward whoever spoke up most in the last meeting.

If you record meetings, an AI note-taker keeps transcripts searchable, which helps when you need to find the exact week a decision was made. Treat the transcript as raw material. The note you write afterward is what goes in the review.

Start in February

The best time to start next year’s review is the month after this year’s ends. February works because goals are fresh, the last review conversation is still in memory, and you can decide what to track before the year gets busy.

One running note per direct report

Give each direct report a single living note instead of comments spread across email, chat, and meeting docs. Keep it short enough that you’ll still be updating it in August. Each entry covers four things.

• The date and context, meaning the project, client, or meeting
• What the person did, described as behavior you saw rather than intent you assume
• The effect on the work, such as a blocker removed, a decision improved, or a risk created
• The follow-up, whether that’s a next step, support they need, or the expectation going forward

Turn 1-on-1 notes into evidence

Your 1-on-1 notes are often the only reliable record of decisions, promises, and coaching. After each check-in, take two minutes to turn the raw notes into one or two lines that could support a sentence in the review. If your notes start out messy, these ChatGPT prompts for 1-on-1 meeting notes help clean them up while the meeting is still fresh.

“Talked about being more responsive” won’t survive the review. “On April 8, agreed to answer client messages the same business day, then met that standard for the next three weeks” will. The first is a memory, and the second is something you can point to.

A year on the calendar

The rhythm doesn’t need to be elaborate. It just needs to exist. Here’s one way to spread the work so nothing piles up in the fourth quarter.

With this rhythm in place, the form arrives and you’re assembling a record rather than drafting from a blank page. That’s the difference between a review that explains the year and one that tries to reconstruct it.

Read the Self-Evaluation First

If your cycle includes a self-evaluation, read it before you write a word of your own review. It’s the employee’s version of the year, and the places where it differs from yours are often the most useful part of the whole packet.

Go through it with your notes open. Mark where their account matches yours. Flag anything they claim that you have no record of, and anything in your notes that they left out. Each gap becomes a question for the conversation. An employee who lists a win you never saw may well be right, and you’d rather learn that before the meeting than during it.

Expect the self-evaluation to arrive with some AI help, because plenty of employees draft them that way now. That’s fine. Our guides to writing self-evaluations with ChatGPT, Claude, Copilot, Gemini, and Notion AI all push employees toward dated examples over polished adjectives, and you can hold the self-evaluation to that same standard. For a sense of what a strong one looks like, these self-evaluation examples show the difference between specific and vague across several situations.

AI can help you compare the two documents. Ask it to line up the employee’s claims against your notes and list where they agree, where they differ, and where one side has no example to point to. Don’t let it smooth over the gaps or guess at why someone wrote what they wrote. Those differences are yours to talk through in person.

Adding 360 and Peer Input

Your view of an employee is partial, sometimes badly so. You see the meetings you attend and the work that reaches your desk. Peers, stakeholders, and direct reports see the rest, which is why outside input can sharpen a review when you collect it carefully.

Ask a small number of people who worked closely with the employee. Three to five is plenty. Ask about specific work instead of general impressions. “How did the vendor migration go from your side?” gets you something you can use, while “What do you think of Dana?” gets you a vibe. If you want help writing questions that draw out specifics, these ChatGPT prompts for 360 feedback are a good starting point.

For anyone who manages people, add input from their own team. Upward feedback is often the clearest view of how someone leads day to day, and it tends to surface things their manager never sees.

Treat peer input as one more source, weighed alongside your own notes. A single harsh comment shouldn’t anchor the review, and neither should a single glowing one. Look for patterns that show up across several people or that match what you’ve seen yourself. AI is good at this sorting step, since it can group comments by theme and flag which ones repeat. Strip names before you paste anything in, and check your company’s policy on putting feedback into AI tools at all.

How to Write a Performance Review That Holds Up

A performance review is a judgment you make and then explain. Your job is to turn what you’ve collected into a clear call, keep that call tied to the job and the review period, and make sure good writing isn’t covering for thin substance. Every sentence should tell the reader what happened, in what context, and with what effect.

Behavior, context, effect

The cleanest review language follows that order. Start with what the person did, add the situation, then state the result. Labels like “strong communicator” or “needs more ownership” hide the actual work, and they give the employee nothing to repeat or fix.

A weak sentence says “handles customers well.” A stronger one says “On two late-stage client escalations in Q2, responded within the agreed window, brought support in early, and kept both accounts on track for renewal.” The second version is specific and observable, and it tells the employee exactly what to keep doing.

If you get stuck on wording, these performance review phrases are written in a person’s voice, so they don’t read like a template. Use them as a starting point and swap in your own examples.

Keep the structure simple

Most reviews work best in four parts, which are a summary, strengths, growth areas, and goals. It’s a basic format, but it’s easy to defend and easy to talk through, and it keeps the main point from getting buried under anecdotes.

Say the main thing in the summary. If the employee did excellent work, say so plainly and show why. If they struggled, say that just as directly and describe the gap between what was expected and what happened. A few clear paragraphs usually do more than a page of careful hedging.

Before you finalize, read the draft for language that judges personality or guesses at attitude. Words like “abrasive,” “lacks passion,” or “not a culture fit” say more about the writer than the work. Replace each one with the behavior that prompted it, or cut it.

Where AI Helps With the Draft

AI earns its place at the drafting stage. It can turn a year of rough notes into an organized first draft in minutes, and it’s good at catching vague phrasing. Things go wrong when you hand it the judgment along with the notes.

Each of the main tools has a lane. ChatGPT is quick at turning scattered notes into a clean narrative, and our guide to ChatGPT for performance reviews walks through that workflow. Claude tends to keep the tone steadier across a long draft, which helps when the review includes hard feedback. If your notes, email, and meetings live in Microsoft 365, Copilot can pull the relevant threads together without you hunting for them. Any of the three will speed up drafting. None of them should own the conclusion.

A workflow that keeps you in charge

1. Give it only notes you trust, meaning dated, first-hand examples with impressions left out.
2. Ask it to sort the evidence into summary, strengths, and growth areas, and leave the rating out of the request.
3. Check every sentence against your notes and cut anything you can’t trace back.
4. Remove guesses about attitude, personality labels, and anything that reads as emotional speculation.
5. Rewrite the final version in your own voice, so the employee hears you in it.

Keep ratings, pay decisions, and anything with legal weight out of the AI tool entirely. Those belong to you and your HR process. Check what your company allows before you paste employee information into any AI tool, and strip names and identifying details whenever you can.

Three prompts worth keeping

Here are my dated notes on one employee from this review period. Group them under summary, strengths, and growth areas. Don't add anything that isn't in the notes, and tell me which sections have thin evidence.

Here's my draft review. Flag any sentence that describes personality or attitude instead of behavior, and any claim that isn't backed by a specific example. Suggest a rewrite for each one.

Here's the employee's self-evaluation and my notes for the same period. List where we agree, where we differ, and where either of us makes a claim without an example. Don't speculate about why we differ.

For more ready-to-use options, see these ChatGPT prompts for performance reviews and Claude prompts for performance reviews.

Setting Goals for the Next Cycle

The goals at the end of a review are what you’ll be measuring next year, so they deserve more than the last five minutes of writing time. Keep them short and tied to work the employee owns. For stable, measurable work, outcome goals fit well. For new, complex, or shifting work, learning goals usually make more sense, because a rigid target can punish someone for adapting.

The CIPD’s evidence review found that goal setting improves performance, and that the effect is stronger when employees can track their progress during the cycle instead of waiting until the end. Goals and feedback should keep feeding each other through the year.

Pick the right kind of goal

If the work is well understood, a direct outcome goal is fine. If it involves a new system, a new market, or a lot of moving parts, set a learning goal first. Someone moving into an unfamiliar customer segment might spend the first half of the year building knowledge before you put a hard number on them. A target can look precise and still be pointless if the person can’t influence the result.

For wording you can adapt, these performance review goals examples name an outcome, an owner, and a check-in date in each one. If the review surfaced a skill gap or a growth path, connect the goals to an employee development plan so the support gets written down too.

Let AI summarize progress

AI is useful when you have months of notes and project updates to sort through. It can summarize progress against each goal, point to where work stalled, and show which goals came up most often in your 1-on-1s. That saves real time. Whether the goal was met is still your call.

Before the review closes, write each new goal in one sentence, confirm the employee can influence the result, and put the first check-in on the calendar now. A goal nobody looks at until next October won’t shape anything.

Special Cases

Not every review fits the neat annual pattern. In each of these, match how firm your conclusions are to how much you can show.

Underperformance and PIPs

When performance is off, keep the review concrete. Describe the expectation, the gap, the support you offered, and what happens if the gap continues. Skip the personality language and the vague disappointment. If a performance improvement plan is in place or likely, the review should line up with it, and the wording should match what HR has on file. These ChatGPT prompts for writing PIPs help keep that language factual. AI can tighten the wording and strip out emotional overreach, but it shouldn’t draw disciplinary conclusions or soften how serious the issue is.

Mid-year reviews

A mid-year review is a progress check, so keep it shorter and more specific. Look at where each goal stands, what’s blocking it, and whether priorities have changed enough to rewrite it. Skip the full summary of strengths unless something has shifted. Its main job is making sure the year-end review holds no surprises. Our guide to AI for mid-year reviews covers a lighter version of this workflow.

A new hire’s first review

Someone three or four months in can’t be measured against a full year. Review their ramp-up instead. How quickly are they learning the work, how are they building relationships, and are they getting the support they need? Say clearly which expectations apply now and which ones start next cycle. Comparing a new hire to a three-year veteran tells you nothing useful.

Visibility versus contribution

This is the biggest trap in hybrid work. Someone who’s visible in every meeting can look stronger than someone who delivers quietly in the background. Before you finalize, ask yourself which examples show outcomes, which show coordination, and which only show who was in the room. Deloitte’s 2025 research found that most managers and workers don’t trust their organization’s performance process, and uneven visibility is a big part of the reason.

One useful habit is a quick calibration with another leader before the review goes final. Ask them whether your examples would hold up if the employee challenged the rating. If the answer is no, go back to your notes.

Running the Review Conversation

The written review is half the job. The conversation is where it either lands or doesn’t.

Send the review a day ahead if your process allows it. People take hard feedback better when they’ve had time to read it alone, and the meeting becomes a discussion instead of a live reading. If you can’t share it in advance, give them a few quiet minutes with it at the start.

Open with the summary in your own words. Then go through strengths and growth areas with the examples behind each one, and ask for their view before you move on. Leave real time for questions. A review meeting that runs entirely in one direction usually means the hard parts haven’t sunk in.

For conversations you expect to be tense, rehearse the difficult part out loud first. These prompts for difficult employee conversations let you practice with AI playing the employee, which takes some of the edge off your first attempt.

End with agreement on what happens next. Confirm the goals, the support you’ll provide, and the date of the first check-in.

When the Employee Disagrees

Some employees will disagree with part of their review, and a few will disagree with all of it. That’s normal, and it doesn’t mean the review failed.

Listen first. Ask which parts they see differently and what they’d point to. If they bring examples you didn’t have, take them seriously and be willing to revise. A review that changes when new facts appear is more credible than one that never moves. If they disagree without anything new, explain your reasoning calmly, walk through the examples, and let the disagreement stand on the record.

Many companies let employees add a written response, so encourage that if they want it. Our guide on what to do when an employee disagrees with their review covers the manager’s side in more detail. It also helps to see the other chair, and this piece on how to respond to a performance review shows what a constructive reply from the employee looks like.

After the Review

The review meeting is where next year’s record starts. Within a week, send a short written summary of what you agreed on, including goals, support, and check-in dates. Then open a fresh running note.

Use your first few 1-on-1s to check whether the goals still make sense now that the work has started. If the review flagged a growth area, put something specific on the calendar, such as a stretch project, a training, or time pairing with someone who’s strong in that skill. The employees who get the most out of a review are the ones who watch their manager follow through on it.

Bottom Line

A good performance review comes down to the notes you keep, the self-evaluation you read closely, and the conversation you prepare for. Start a running note for each direct report this month, turn your 1-on-1s into dated examples, and let AI handle the sorting and tightening while you keep the judgment. When the form shows up next fall, most of the work will already be done.

Frequently Asked Questions