ChatGPT Prompts for HR (Job Posts and Policies)

ChatGPT prompts for HR work when you feed a role brief, a competency list, and a fairness fence, then demand one artifact: job post, screening notes, interview kit, or policy draft. You leave with ROLE / TASK / FORMAT shapes for GPT-5.5 Instant, GPT-5.6, Claude Sonnet 5, or Gemini 3.5 Flash, plus rules that block biased wording, protected-class guesses, and invented legal claims. This guide is for HR generalists and recruiters who write hiring materials. Treat the output as a draft. A human must review before you publish a post, score a candidate, or ship a handbook. PromptMake /text can turn a rough HR ask into a labeled scaffold at https://promptmake.net/text before you paste into your chat model.

Who ChatGPT prompts for HR help

You turn messy manager notes into a clean artifact without guessing who a candidate is or inventing a statute. The patterns fit recruiters who turn a hiring-manager brain dump into a job post, and HR generalists who build structured screening notes from a resume against must-haves. HRBPs who need a first-pass interview kit so each candidate at a stage hears the same questions land here too, as do people-ops writers who draft handbook language from a current policy and a change request.

Skip candidate-side mock interviews; that work lives in the interview preparation guide. Resume rewrites and cover letters have their own libraries. This article stays on the employer side: public job posts, internal screening notes, interviewer kits, and policy drafts. You supply ROLE_PACK, COMPETENCY_LIST, audience, and the fairness fence. The model proposes section order, tighter verbs, and consistent scorecard columns. You reject anything that infers a protected class, invents a pay band, or writes a hire / no-hire line as if it were a decision.

Company AI policy, privacy rules, and employment law still govern what you paste and what you publish. Check your counsel and your vendor rules before you drop candidate files into a consumer chat. Redact names, emails, phone numbers, photos, and unique facts that would identify a person. Use Candidate A / Role 12 labels. Keep the official file in your ATS, not in the chat history.

Core prompt pattern for HR writing

Strong ChatGPT prompts for HR give four inputs before any tone request: the role pack, the competency list, the deliverable, and the fairness fence. The role pack holds the job title, team, location, employment type, pay band you may publish, and essential functions you can defend with the hiring manager. The competency list holds skills and behaviors tied to the work, split into must-have and nice-to-have. The deliverable names the artifact: job post, screening notes, interview kit, or policy draft. The fairness fence forbids protected-class guesses, gendered wording, age proxies, culture-fit scores, and invented legal claims.

Paste those four blocks near the top. Put format rules next. Repeat the fence at the end so it survives a long paste. GPT-5.5 Instant, GPT-5.6, Claude Sonnet 5, and Gemini 3.5 Flash follow labeled blocks. Vague "write a job description" prompts produce generic essays because the model has no competency list and no banned-language list. Aim for one artifact type per thread. Mixing a public job post and an internal scorecard in the same pass leaks screening language into copy candidates will read. Keep ROLE_PACK and COMPETENCY_LIST in a local file. Feed them into each prompt that touches the draft.

Role, task, format skeleton

Copy this skeleton and fill the brackets with your material:

ROLE: You are an HR drafting assistant for [company type]. You draft from ROLE_PACK and COMPETENCY_LIST only. You never infer race, color, religion, sex, gender, pregnancy, sexual orientation, gender identity, national origin, age, disability, genetic information, veteran status, or marital status. You never invent pay bands, visa rules, statutes, or headcount. You do not make a hire or no-hire decision. You produce a draft for HR to review.

TASK: Turn ROLE_PACK and COMPETENCY_LIST into a [job post | screening notes | interview kit | policy draft] for [audience]. Keep every requirement tied to COMPETENCY_LIST. Keep every fact tied to ROLE_PACK. Mark gaps with [UNKNOWN]. Mark any line that needs a human call with [HR REVIEW]. Mark any legal claim with [LEGAL REVIEW].

FORMAT: Use the section headers listed under OUTPUT_SHAPE. Short sentences. Must-have and nice-to-have in separate lists. If a fact is missing, write [UNKNOWN] and do not guess. Ban culture-fit scores and personality labels.

AUDIENCE: [public candidates, hiring manager, interview panel, or employees]

ROLE_PACK: [paste title, team, location, employment type, approved pay band or [UNKNOWN: pay], essential functions, schedule, and open questions marked ?]

COMPETENCY_LIST: [paste must-have skills and behaviors, nice-to-have extras, and how you will see them on the job]

SOURCE_POLICY: [for handbook drafts only: paste current policy text and the change request]

OUTPUT_SHAPE: [list required sections or columns]

RULES: No "ninja," "rockstar," "digital native," "young and energetic," or "culture fit." No gendered job titles. If a criterion is missing, write [UNKNOWN]. Never invent PII. This output is a draft for human review, not a hiring decision and not legal advice.

REMINDER: Never infer protected classes. Never invent pay, visa, or statute claims. Use [UNKNOWN], [HR REVIEW], and [LEGAL REVIEW].

That reminder line stops confident fiction. Models love tidy salary ranges and a punchy culture paragraph. Your rule forces a gap marker you can fill from compensation, immigration, or counsel after you check the source.

Fairness and bias fences you paste every time

Build a short FAIRNESS_FENCE block you paste under RULES every time:

FAIRNESS_FENCE: Do not infer or mention protected classes from names, schools, photos, voice, gaps, or resume clues. Do not use age proxies (digital native, recent graduate as a must-have unless ROLE_PACK names a true internship). Do not use gendered titles or coded words (ninja, rockstar, culture fit, native speaker unless the job requires a named language skill you can test). Score only against COMPETENCY_LIST. Use the same questions for all candidates at this stage. Prefer [UNKNOWN] over a guess. Prefer [HR REVIEW] over a hire or no-hire line. This draft needs human review and is not a legal opinion.

Keep the review tokens distinct. [UNKNOWN] means the pack has a hole: missing pay band, missing essential function, missing current policy text. [HR REVIEW] means the text states a screening score, a ranking, or a recommendation a human must own. [LEGAL REVIEW] means the text mentions a statute, an EEO sentence, leave eligibility, or a claim about what the company "must" do under law.

For public job posts, paste the EEO statement your company already uses. Add: "Copy the EEO_STATEMENT verbatim. If EEO_STATEMENT is empty, write [UNKNOWN: EEO] and do not invent legal language." For screening notes, add: "Do not comment on name, school prestige, employment gaps, or accent. If those appear in SOURCE, ignore them unless COMPETENCY_LIST names a job-related license or credential." For interview kits, require the same question list and the same scoring anchors for every candidate at that stage.

AI prompts for HR job posts, screening notes, and kits

Draft in this order when the req is new: job post, then screening notes, then interview kit. Policy drafts sit on a separate thread so handbook language does not leak into candidate copy. You lock must-haves in the public post before you score anyone. Screening notes then map a redacted resume to those must-haves. An interview kit reuses the same competencies so the panel does not invent a new bar mid-loop. Reuse ROLE_PACK, COMPETENCY_LIST, and FAIRNESS_FENCE. Change only TASK and OUTPUT_SHAPE. After each draft, run a short fairness audit before you move on.

Keep audience labels tight. A public post uses plain requirements and a pay line you already approved. Internal screening notes may use score columns, but those notes stay out of the public post. Ask the model for the public post first. Convert to a scorecard in a second pass only after you lock the must-haves with the hiring manager.

Job posts and screening notes

Job post pattern:

TASK: From ROLE_PACK and COMPETENCY_LIST, draft a job post under [word cap]. Sections: Title, Team and location, What you will do (essential functions), Must-have, Nice-to-have, Pay and benefits (only facts in ROLE_PACK), How to apply, EEO_STATEMENT (verbatim or [UNKNOWN: EEO]). Split must-have and nice-to-have. Ban years-of-experience as a proxy unless ROLE_PACK states a license that needs a year count. Ban "native speaker" unless COMPETENCY_LIST names a language test. Mark missing pay as [UNKNOWN: pay].

Job post audit pattern: "TASK: Audit JOB_POST against ROLE_PACK, COMPETENCY_LIST, and FAIRNESS_FENCE. List every requirement, pay claim, visa claim, and personality phrase. Reply Keep, Soften, or Remove. Soften means the line overreaches the pack or reads as a proxy for age, gender, or disability. Propose a rewrite. Never add a new requirement."

Screening notes pattern:

TASK: From REDACTED_RESUME and COMPETENCY_LIST, draft screening notes. Columns: Competency (must-have only), Evidence from resume (quote or paraphrase the line), Score 1-3 using SCORE_ANCHORS, Notes, Flag ([UNKNOWN] if no evidence, [HR REVIEW] if you would recommend next step). Do not score nice-to-haves as knockouts. Do not mention name, school rank, gaps, or photos. Do not output hire or no-hire. Cap Notes at 25 words per row.

SCORE_ANCHORS belong in the prompt: 1 = no evidence of the competency, 2 = partial evidence, 3 = clear evidence tied to the work. You write those anchors. The model fills the table. A human still decides who advances.

Interview kits and policy drafts

Interview kit pattern:

TASK: From COMPETENCY_LIST and INTERVIEW_STAGE, draft an interview kit. Include: Purpose (one paragraph), Competencies in scope, Question list (same order for all candidates), Scoring anchors per question (1 / 2 / 3), Interviewer do and do-not list, Time box. Cap at [N] questions. Each question must map to one must-have competency. Ban brainteasers, illegal topics (age, family plans, health, religion, national origin), and "tell me about yourself" as a scored item. Mark any question that needs a fact missing from ROLE_PACK with [UNKNOWN]. Add a header: Draft for HR review. Not a hiring decision.

Interviewer do-not list should name the fence in plain words: do not ask about family, health, age, accent, or social media. Do not score "polish" or "likability." Score the answer against the anchor. Run a second pass: "TASK: Audit KIT against FAIRNESS_FENCE. Flag any question that could elicit protected-class information. Rewrite or remove. Never add a new competency."

Policy draft pattern:

TASK: From SOURCE_POLICY and CHANGE_REQUEST, draft a handbook section under [word cap]. Sections: Purpose, Who it covers, Rules, Examples, How to ask a question, Related policies named in SOURCE_POLICY. Use only leave types, dollar amounts, and waiting periods present in SOURCE_POLICY. If CHANGE_REQUEST wants a new legal claim, write [LEGAL REVIEW] and [UNKNOWN: statute] instead of inventing a code section. Tone: plain, second person to employees. Ban "the company may terminate at any time for any reason" unless that sentence already appears in SOURCE_POLICY. Header: Draft for HR and employment counsel. Not legal advice.

Do not ask the model to "make this policy bulletproof" or to hide a rule from employees. Ask for a draft counsel can edit. HR and a lawyer pick the language and record the business call.

Step-by-step HR prompt workflow

Use one chat thread per req and artifact: Role 12 job post, Role 12 screening, Role 12 interview kit. If you dump every open req into one long thread, you mix must-haves and invent shared pay bands. Keep a stable ROLE_PACK while you swap only TASK and OUTPUT_SHAPE. You spend free ChatGPT, Claude, Gemini, or PromptMake runs on structure, then HR time on fairness review and manager sign-off. Policy work gets its own thread with SOURCE_POLICY, so a PTO rewrite cannot bleed into a job post.

Keep ROLE_PACK in a local file: title, team, location, employment type, approved pay, essential functions, and open questions. Keep COMPETENCY_LIST as a separate list you agreed with the hiring manager. Keep candidate files in the ATS. The model should never be your system of record for people or pay. If your company forbids consumer tools for candidate data, stop and use the approved vendor instead. The prompt shapes still apply inside that vendor.

Step 1: Build the role pack and competency list

List what you may say. Example: "Role 12. Support Lead. Team of five. Hybrid Austin. Full-time. Pay band $85k-$105k (comp approved 2026-07). Essential functions: ticket triage, weekly 1:1 coaching, weekend on-call. Must-have: Zendesk admin, written English for tickets, coaching. Nice-to-have: Salesforce. Ban: ninja, rockstar, culture fit, years as a knockout. EEO_STATEMENT: [paste yours]." Ugly notes beat a polished fake post.

Mark every pay, visa, and legal line as typed-from-source or absent. In the prompt, tell the model to keep absent items as [UNKNOWN]. Guessing "$90k-$110k" when you mean "comp has not approved a band" creates a public post you cannot honor and a possible pay-transparency problem in places that require a range.

Step 2: Outline, then draft one artifact

Name the artifact and audience before you paste long notes. Ask for an outline first when the post is new:

TASK: From GOAL and AUDIENCE, propose an outline with headers only. Do not draft body text yet. Flag any header that needs facts missing from ROLE_PACK or COMPETENCY_LIST.

Use that outline to decide what to paste next. Partial facts get [UNKNOWN] subsections. Missing pay stays out of the public post until compensation signs off. Then run the full skeleton with OUTPUT_SHAPE matching the outline you approved. For screening, redact the resume first: Candidate A, no photo, no address, no dates of birth. Paste only the work history and skills the scorecard needs.

Step 3: Fairness audit, then human review

Take the draft into a second message:

TASK: Audit DRAFT against ROLE_PACK, COMPETENCY_LIST, and FAIRNESS_FENCE. For each requirement, pay claim, visa claim, score, and personality phrase, reply Keep, Soften, or Remove. Soften means the claim overreaches the pack or works as a proxy. Propose a safer rewrite. Never add new requirements or a hire decision.

Optional scaffold: open https://promptmake.net/text, describe "HR prompts for job posts, screening notes, interview kits, and policy drafts with fairness fences, [UNKNOWN] gaps, and [HR REVIEW] flags," generate once, then paste your ROLE_PACK into the returned structure. Guests get about three runs per day; free accounts get about five. Use a run to shape the prompt, then finish in your chat model.

Save the winning prompt next to the req ID and model label so the next post reuses the same wrapper. Publish or advance a candidate only after a human checks every line against the pack and owns the decision.

Mistakes that wreck HR drafts

Mistake 1: Asking the model to "write the job description" with no ROLE_PACK or COMPETENCY_LIST. The model invents years, tools, and a culture paragraph that fail the first manager review. Paste controlled notes instead.

Mistake 2: Allowing culture-fit scores, personality labels, and age proxies. If you did not name a job-related competency, ban it in FORMAT. "Digital native" and "young team" create legal and fairness risk. Require [UNKNOWN] or a rewrite.

Mistake 3: Treating the draft as a hiring decision. Put "This is a draft for HR review. Not a hire or no-hire decision." in ROLE and in the scorecard header the model must output. A human still advances or rejects.

Mistake 4: Using GPT-5.5 Instant or Gemini 3.5 Flash as the final fairness checker. Fast models fit outlines and first-pass posts. Route the fairness audit to GPT-5.6, Claude Opus 5, or Gemini 3.1 Pro, then still read the draft yourself.

Mistake 5: Pasting unredacted resumes, offer letters, or employee files into a consumer chat without your company's AI policy check. Redact. Use candidate labels. Keep sensitive fields offline.

Mistake 6: One mega-prompt that asks for a job post, a scorecard, and a rejection email together. Split artifacts. Reuse ROLE_PACK; change TASK and FORMAT.

Mistake 7: Asking the model to invent an EEO statement, a statute, or a "we do not sponsor" rule you did not supply. Paste SOURCE_POLICY or write [UNKNOWN]. Counsel owns legal lines.

Mistake 8: Skipping the audit after a long paste. Narrative posts hide "native speaker" and "must be able to lift" lines you never approved as essential functions. Prompt for Keep / Soften / Remove against FAIRNESS_FENCE every time.

Model notes for HR writing (mid-2026)

ChatGPT often defaults to GPT-5.5 Instant for fast chat. Instant fits job-post outlines, first-pass screening tables, and interview-kit headers when you already locked COMPETENCY_LIST. Keep prompts short: ROLE, TASK, FORMAT, ROLE_PACK, COMPETENCY_LIST, RULES. Skip long chain-of-thought slogans.

GPT-5.6 (Sol in API naming as of mid-2026) fits harder edit passes: fairness audits against FAIRNESS_FENCE, contradiction checks between must-haves and nice-to-haves, and policy drafts that must not invent statutes. Give goal, constraints, and format. Drop "think step by step" padding on reasoning-class models.

Claude Sonnet 5 handles long ROLE_PACK pastes and tidy scorecard tables well. Claude Opus 5 fits careful audits when the kit must not invent questions or scoring tricks. Claude Fable 5 is the widely released top tier when your workspace offers it; check your plan. Haiku 4.5 fits short outline reshuffles when latency matters more than a deep audit.

Gemini 3.5 Flash fits volume work: many outline passes and screening tables from an updated COMPETENCY_LIST. Gemini 3.1 Pro fits hard reasoning over long context when you paste a thick policy pack and need conflict flags against SOURCE_POLICY. Hedge on exact menu names in each vendor UI. They shift. Re-check the model picker when you open a new thread.

Prompting split that holds: Instant and Flash get RTF plus a short sample of your house job-post format when you need matching structure. GPT-5.6, Opus 5, and Gemini 3.1 Pro get goal + constraints + format, with an explicit "never infer" rule, a [UNKNOWN] token, and [HR REVIEW] / [LEGAL REVIEW] flags. All need your ROLE_PACK in the message. None replace a human fairness check or employment counsel before anyone relies on the draft.

Build ChatGPT prompts for HR with PromptMake

Write the rough ask in plain words: artifact type (job post, screening notes, interview kit, policy draft), fairness fence, and whether you need a Keep / Soften / Remove audit. Open https://promptmake.net/text and generate a structured prompt once. Expect labeled sections you can fill with ROLE_PACK and COMPETENCY_LIST.

Edit titles, pay bands, and EEO language yourself. PromptMake cannot know your req or your handbook. Paste the filled prompt into Instant or Flash for drafts, or GPT-5.6 / Opus 5 / Gemini 3.1 Pro for audits. Keep free-tier runs for scaffolding, not five synonym retries of the same weak ask.

Workflow that sticks: ROLE_PACK + COMPETENCY_LIST → PromptMake scaffold → fill redacted facts → fast model draft → reasoning model fairness audit → human HR review → publish or advance only after that review. Store one template per artifact so you do not rewrite ROLE and RULES from scratch each req.

FAQ

What are the best ChatGPT prompts for HR in 2026?

The best ChatGPT prompts for HR lead with ROLE and a fairness fence, paste a ROLE_PACK and a COMPETENCY_LIST you agreed with the hiring manager, then demand FORMAT for one artifact: job post, screening notes, interview kit, or policy draft. Add a second audit prompt that marks Keep, Soften, or Remove against those packs. Match GPT-5.5 Instant or Gemini 3.5 Flash for drafts and GPT-5.6, Claude Opus 5, or Gemini 3.1 Pro for fairness audits, then still read the draft yourself before anyone sees it.

Can ChatGPT write a job description from scratch?

The model can draft structure and wording from a role pack and competencies you supply. It should not invent pay bands, years-of-experience knockouts, or a culture paragraph you cannot defend. Start from ROLE_PACK notes you wrote offline, and treat blank-slate "write our JD" prompts as high risk for fiction that fails a manager review. A human still owns the posted version.

How do I keep AI prompts for HR free of biased language?

State the ban in ROLE and again in FAIRNESS_FENCE plus a REMINDER line. Forbid protected-class guesses, age proxies, gendered titles, and culture-fit scores, and require [UNKNOWN] when a stronger claim wants a fact you did not give. Follow with an audit prompt that compares DRAFT to COMPETENCY_LIST line by line. Then a human still reads for proxies the model missed.

Are AI prompts for HR the same as interview prep prompts?

AI prompts for HR target employer-side artifacts: job posts, screening notes, interviewer kits, and policy drafts with fairness fences and [HR REVIEW] on scores. Interview preparation prompts target candidates who need mock questions, STAR stories, and thank-you emails. Keep those libraries separate so a candidate rehearsal prompt does not turn into a scorecard, and a scorecard does not invent a hire decision.

Should I use GPT-5.5 Instant or GPT-5.6 for ChatGPT prompts for HR?

Use Instant for job-post outlines, first-pass screening tables, and interview-kit headers from a competency list you pasted. Use GPT-5.6 when you need a careful fairness audit, conflict checks between must-haves and nice-to-haves, or a policy draft that must not invent statutes. Run the same packs through both only when you measure quality for a recurring template.

Can I use an AI score to reject a candidate?

Do not treat an AI score as a reject decision. Put that limit in ROLE and in the scorecard header. Screening notes can map evidence to competencies and mark [HR REVIEW]; they must not output a hire or no-hire fact. Your process, your ATS, and employment rules still require a human decision, so check your company's AI policy before you paste candidate data into a consumer tool.

Can PromptMake help with ChatGPT prompts for HR on the free tier?

Yes. PromptMake /text turns a rough HR ask into a labeled prompt you can aim at ChatGPT, Claude, or Gemini. Guests get about three generations per day; registered free users get about five. Fill in your own ROLE_PACK and COMPETENCY_LIST, then paste into Instant, Flash, or a reasoning model for the audit. PromptMake does not make hiring decisions and does not store your ATS as a system of record.