The short answer
The best prompts for HR help with adverts, structured interview guides, onboarding plans, policies and announcements, and never ask the AI to rank or reject people. Keep candidate and employee data out of consumer plans: use a work plan such as ChatGPT Business, Claude Team or Microsoft Copilot with a work account, or anonymise first. The EU AI Act lists AI for recruitment and worker management as high-risk, with those rules applying from 2 December 2027.
How to use these HR prompts safely
HR work is full of writing that AI does well: job adverts, interview guides, policies, onboarding plans and announcements. It is also full of personal data and decisions that affect people's lives, which is why these 12 prompts follow two rules. First, they never ask the AI to judge, rank, screen or reject a person. Second, they work with job requirements, policies and anonymised text, not with names, health details or other personal information.
Both rules matter beyond good practice. The European Commission lists "AI tools for employment, management of workers and access to self-employment (e.g. CV-sorting software for recruitment)" as high-risk under the EU AI Act, and the rules for those uses apply from 2 December 2027, after the AI Omnibus delayed them from August 2026. Emotion recognition in workplaces has been banned since 2 February 2025. Using a chat assistant to write an interview guide is not the same as running an automated hiring system, but the direction is clear: keep a human making every decision about a person, and keep a record of how you decided. Read our explainer on the EU AI Act for more.
Keep personal data out of consumer plans
Do not paste candidate CVs, employee records, health information, grievances or performance reviews into a free or personal AI plan. On the free and personal plans of ChatGPT, Claude and Perplexity, chats can be used to train models unless the user switches that off, and with Gemini on a personal account, activity is used to improve Google's services while Keep Activity is on. Switching training off is not enough for HR data: you still need a lawful basis, your company's approval and a contract that covers the data.
Use your company's work plan instead. ChatGPT Business and Enterprise and Claude Team and Enterprise do not train on your content by default. Microsoft Copilot Chat signed in with a work or school account runs under enterprise data protection, and Gemini in Google Workspace does not use your content for training outside your domain without permission. If you have none of these, anonymise first: remove names, replace small teams with larger groups and strip details that could identify someone. Our guides to private AI for sensitive data and using AI at work safely go further.
Prompts for fair hiring
Write an inclusive job advert, Audit an existing job advert for bias, Build a structured interview guide and Write a scoring rubric for a work sample help you hire on evidence. Structured interviews, where every candidate gets the same job-related questions and is scored against the same anchors, are one of the most useful things you can introduce, and AI makes the guide fast to write. Two warnings. AI can repeat bias from its training data, so always read its output for assumptions about who does a job. And never ask it which candidate to hire: compare scored notes as a panel.
Prompts for onboarding and policies
Draft a 30-60-90 day onboarding plan, Turn a policy into plain English and Draft a new policy for legal review save hours of drafting. The policy prompts tell the AI to keep every rule unchanged and to flag ambiguity, because a friendly rewrite that quietly changes an entitlement is a real risk. The [LEGAL CHECK] marker keeps the AI from stating employment law as fact. Employment law varies by country and changes often, so a qualified person checks the draft.
Prompts for employee communication and listening
Announce a change to employees, Analyse anonymised survey comments, Prepare for a difficult conversation, Create a manager training scenario and Write a fair exit interview guide cover the everyday HR work around people. The survey prompt only works safely on comments that are already anonymised, with small teams merged so nobody can be identified by their team. Tell the AI not to guess authors and check the theme counts against a sample of the raw comments yourself.
How to adapt these prompts
- Add your country and company size. Policies, adverts and interview rules differ between the UK, Germany, the US and elsewhere.
- Paste your values or competency framework into the interview and rubric prompts, so scoring matches how your company already defines good work.
- Save your house style. Put your tone of voice and template headings in a saved project or custom instructions in your approved tool, so every draft starts consistent.
- Keep the human step visible. End hiring and performance prompts with "List what a human must decide here" so the boundary stays clear.
Which AI tool works best for HR
The best tool is the one your company has approved on a work plan, because that decides where employee data can go. If your organisation uses Microsoft 365, a Microsoft Copilot licence works inside Word, Outlook and Teams with your files, and even the included Copilot Chat runs under enterprise data protection, with prompts and responses not used to train the underlying models. For drafting policies and long documents, Claude on a Team or Enterprise plan is our pick, as it keeps careful wording and flags ambiguity well. ChatGPT Business, for 2 to 200 people, is a strong all-round choice, and Gemini suits teams on Google Workspace. See the best AI for HR and recruiting and AI for HR teams for a full comparison.
Prompts to copy
Write an inclusive job advert
Write a job advert for a [job title] in our [team] at [company type and size], based in [location / remote]. The must-have requirements are: [paste, maximum five]. Nice to have: [paste]. Salary range: [range]. Use plain, gender-neutral language, avoid phrases such as "young and dynamic" or "native speaker" that can put off or exclude qualified people, and do not ask for years of experience unless the job truly needs them. Under 400 words. Afterwards, list any requirement that might exclude good candidates without being necessary.
You get: A clear advert plus a check of requirements that may narrow your pool unfairly
Audit an existing job advert for bias
Review this job advert: [paste]. Flag any wording that could discourage applicants because of gender, age, disability, ethnicity, religion, family status or nationality, and explain why in one line each. Flag requirements that look stricter than the job needs. Suggest a neutral rewrite for each flag. Do not change the actual duties.
You get: A line-by-line bias review with neutral rewrites
Build a structured interview guide
Create a structured interview guide for [job title]. Base it only on these job requirements: [paste]. For each requirement, write two behavioural questions, what a strong, adequate and weak answer would show, and a 1 to 4 scoring scale. Every candidate gets the same questions. Do not include any question about age, family plans, health, religion, nationality or other personal characteristics, and add a short note for interviewers on questions to avoid.
You get: A fair, job-related interview guide with scoring anchors
Write a scoring rubric for a work sample
We give candidates for [role] this task: [describe]. Write a scoring rubric with four or five criteria tied to the job requirements: [paste]. For each criterion, describe what a score of 1, 2, 3 and 4 looks like with concrete examples. Add guidance on scoring the work itself rather than writing style or polish that is not part of the job.
You get: A rubric that makes different interviewers score the same way
Draft a 30-60-90 day onboarding plan
Draft a 30-60-90 day onboarding plan for a new [job title] joining [team]. Their first key goals are: [paste]. Include who they should meet in week one (by role, not name), the systems they need access to, one early win they can deliver in the first month and check-in points for their manager. Format it as a table they can tick off.
You get: An onboarding plan the new hire and manager can both use
Turn a policy into plain English
Here is our [parental leave / hybrid working / expenses] policy: [paste, with names removed]. Rewrite it as a one-page employee guide in plain English: who it applies to, what they are entitled to, how to request it and who to ask. Keep every rule exactly as it is in the original and flag anything that is ambiguous instead of guessing what it means.
You get: A readable guide that does not quietly change the rules
Draft a new policy for legal review
Draft a first version of a [policy name] policy for a [size] company in [country]. Cover purpose, scope, the rules, responsibilities, how to raise concerns and review date. Mark every place where local employment law or our works council or union agreement must be checked with [LEGAL CHECK]. Do not state legal requirements as facts.
You get: A structured draft with clear flags for your lawyer
Announce a change to employees
Write an internal announcement about [the change, for example a new benefits provider]. Audience: [all employees / one team]. Explain what is changing, when, why, what employees need to do and where to ask questions. Under 250 words, warm and direct. Then write the five questions employees are most likely to ask, with short answers based only on these facts: [paste].
You get: An announcement plus a ready Q and A
Analyse anonymised survey comments
Here are anonymised free-text comments from our engagement survey, with no names, teams under [ten] people merged and identifying details removed: [paste]. Group them into themes, count how many comments fall into each, quote two representative comments per theme and suggest one action per theme. Do not try to work out who wrote anything.
You get: Themes with counts and quotes, ready for a leadership readout
Prepare for a difficult conversation
I am an HR partner preparing a manager for a conversation about [performance concerns / a complaint / a role change]. The facts, with no names: [paste]. Help me plan: the opening sentence, the three key points, how to respond if the employee is upset or disagrees, what not to say and how to agree next steps. Keep it fair, factual and respectful.
You get: A conversation plan that keeps the meeting calm and fair
Create a manager training scenario
Write a realistic role-play scenario for training new managers on [giving feedback / handling a flexible working request]. Include a short brief for the manager, a separate brief for the employee with their perspective, three moments where the conversation could go wrong and debrief questions for the trainer. Use fictional names and a fictional company.
You get: A ready-to-run training exercise
Write a fair exit interview guide
Create an exit interview guide of ten open questions for employees leaving [company type]. Cover role, manager, growth, pay and benefits, culture and what would have made them stay. Write the questions so they invite honest answers without leading. Add a short script for opening and closing the interview, including how the answers will be used and who will see them. Do not promise more confidentiality than the process can keep.
You get: Exit interview questions that get usable answers