Can AI Write Performance Reviews? What Is Acceptable and What Is Not
Many managers already use AI to help write performance reviews, whether their organisation has a policy or not. The question is not whether AI can write a review, but which parts of a review it should touch.
Short answer
AI can draft performance review text, but it should not write reviews on a manager's behalf. Acceptable uses include summarising evidence the manager provides, improving clarity and structure, checking for biased or vague language and suggesting development actions. Unacceptable uses include generating assessments without evidence, deciding ratings, inventing examples or pasting employee data into unapproved tools. The manager must verify every statement and remain the accountable author.
Key takeaways
- AI as editor and research assistant: acceptable. AI as author and judge: not acceptable.
- Every factual claim in a review must be verified by the manager.
- Employees can usually tell when a review is generic; it damages trust.
- Organisations need a clear policy rather than silent, inconsistent use.
Acceptable and unacceptable uses
| Acceptable | Unacceptable |
|---|---|
| Summarising notes and feedback the manager provides | Generating a review from a name and job title |
| Improving clarity, structure and tone | Inventing examples or achievements |
| Flagging vague or biased language | Deciding or justifying the rating |
| Suggesting development actions | Writing reviews the manager has not read carefully |
| Preparing questions for the conversation | Pasting employee data into unapproved consumer tools |
Why fully AI-written reviews fail
- Inaccuracy: AI may misattribute work, exaggerate or invent details.
- Generic language: reviews lose the specificity that makes feedback useful.
- Loss of trust: employees who sense an AI-written review may conclude their manager did not engage.
- Accountability gaps: managers must be able to explain every statement, especially if a review is challenged.
- Legal exposure: reviews can inform pay, promotion and dismissal decisions and may be disclosed in disputes.
- Privacy: performance data entered into consumer tools may be retained by the provider.
A policy framework for AI-assisted reviews
- Approved tools only: specify which tools may process performance data.
- Evidence first: AI may only work from evidence the manager supplies.
- Manager authorship: the manager reviews, edits and owns every word.
- Ratings are human: AI must not propose or justify ratings.
- Verification: managers confirm every factual statement.
- Transparency: employees are informed that AI may assist with drafting.
- Training: managers learn good prompts and common AI errors.
Include these rules in your AI policy for employees.
A quick self-check for managers
- Could I explain every sentence in this review to the employee face to face?
- Is every example real and accurately described?
- Does it sound like me, and like this specific person?
- Would I be comfortable if this review were read in a tribunal or by a works council?
For the full manager workflow, see AI in performance reviews.
Related guides
- How AI Is Used in Performance Reviews: A Practical Guide for Managers and HR
A manager workflow for AI-assisted reviews, with prompts, bias checks and safeguards.
- AI Policy for Employees: Why You Need One and What It Should Cover
Why an employee AI policy is essential, what it covers and who owns it.
- How HR Can Use Generative AI: A Practical Operating Model
The tasks to hand to generative AI, a draft-review-own workflow, guardrails and team skills.
- Risks of AI in Performance Appraisal: What HR Must Guard Against
Eight risks of AI in appraisal, a risk matrix and specific safeguards.
Frequently asked questions
Is it OK to use ChatGPT to write a performance review?
Using an approved AI tool to summarise your own evidence, improve wording and check for bias is generally acceptable. Letting it write the review without your evidence, or entering employee data into an unapproved consumer tool, is not.
Can employees tell if a review was written by AI?
Often, yes. AI-written reviews tend to be generic and lack specific examples, which can make employees feel their manager did not engage.
Should companies ban AI for performance reviews?
Bans are hard to enforce and forgo real benefits. A clear policy on approved tools, evidence-based use, manager authorship and human ratings is usually more effective.
Who is responsible for an AI-assisted performance review?
The manager who signs it off remains fully responsible for its content and the rating, regardless of how it was drafted.