🚀 Tested on GPT-6.1 / GPT-4o, Claude Sonnet 5.5 & Gemini 4 Argon Models verified: October 2026

AI Prompts for HR & Recruiting — Free Working Generators

⚡ Quick Answer

AI prompts for HR work best when they are structured: a defined role, the exact inputs supplied, numbered instructions, and rules that forbid invented requirements. These five free generators cover the recruiting lifecycle — job descriptions, Boolean sourcing, outreach, STAR scorecards, and onboarding — and every one runs entirely in your browser.

⚡ AI Prompts for HR & Recruiting — Free Working Generators

Configure your facts below to generate a production-ready, constraint-locked prompt.

100% Client-Side

What this hub is

Most "AI prompts for HR" pages give you a numbered list: fifty prompts, a hundred prompts, each one a sentence or two long, none of them tested against the realities of hiring. You copy one, paste it into a chatbot, get a generic job description full of requirements nobody agreed to, and spend longer fixing it than writing it yourself would have taken.

This hub works differently. Each page below is a working generator, not a list item. You enter the details of your role, your candidates, or your process; the generator assembles a complete, production-grade prompt — with the guardrails recruiters actually need already built in — and hands you the finished prompt to run in GPT-6.1 / GPT-4o, Claude Sonnet 5.5, or Gemini 4 Argon. Underneath every generator you will also find the full prompt template in the open, so you can read exactly what it does, adapt it, and reuse it without the tool.

There are five generators, covering the recruiting lifecycle end to end:

  1. Job Description Generator — a complete, inclusive job description from your intake notes, plus a bias-and-accuracy report that flags anything the AI added or worded badly.
  2. Boolean Sourcing String Generator — validated Boolean strings for LinkedIn Recruiter, the LinkedIn search bar, Google X-ray, and GitHub, checked for balanced parentheses and dangling operators before you run them.
  3. Candidate Outreach Generator — three personalised outreach variants (curiosity-led, outcome-led, peer-referral) plus a follow-up bump, each inside the platform's character budget.
  4. STAR Interview Scorecard Generator — a weighted scorecard whose weights are forced to sum to 100%, with observable 1/3/5 rating anchors and a debrief sheet.
  5. 30-60-90 Onboarding Plan Generator — three separate checklists (new hire, manager, People Ops) with check-in questions and risk signals for each phase.

Why HR prompts fail — and what these generators do about it

Generic HR prompts fail in predictable ways, and in hiring those failures are not just annoying — they can be discriminatory or unlawful:

How it works

  1. Pick your task. Choose one of the five generators above — or stay here and use the general HR prompt builder embedded at the top of this page.
  2. Enter only what you know. Each generator asks for the inputs that actually change the output: the role, the genuine must-haves, the team context. Leave a field empty and the generated prompt tells the model to flag the gap — never to fill it with invention.
  3. Copy the generated prompt. Run it in GPT-6.1 / GPT-4o, Claude Sonnet 5.5, or Gemini 4 Argon. Every template ends with a "Best model for this" recommendation, because the models are not interchangeable for HR work — long, instruction-dense drafting and nuanced language review behave differently across them.
  4. Run the audit pass. Each template's output includes a short self-audit: added claims, coded phrases, unverified facts. Read it before anything leaves your desk.
  5. Keep a human decision. These prompts draft, structure, and check. They do not decide who gets hired, and every template says so.

The master HR prompt (the pattern every generator follows)

Every generator on this hub is a specialisation of one master pattern. If you only take one prompt from this page, take this one — it turns any one-line HR request into a structured, guarded prompt:

PROMPT TEMPLATE 
You are an experienced HR specialist and recruiter supporting a hiring team.

CONTEXT — use only what is supplied here:
- Task: [describe the HR task in one sentence]
- Role / team: [job title, level, department]
- Key facts I am giving you: [paste the facts, notes, or data the output must be based on]
- Audience: [who will read the output — candidates, hiring managers, employees]

INSTRUCTIONS:
1. Complete the task using ONLY the facts supplied above. Do not add requirements, qualifications, figures, dates, company facts, or candidate details that are not in the context.
2. If information you need is missing, do not guess. Complete what you can and list the missing items at the end under "Questions before this is final".
3. Write in gender-neutral, age-neutral language. Avoid "rockstar", "ninja", "guru", "high-energy", "digital native", "young and energetic", and "culture fit" — describe the actual work and the actual behaviours instead.
4. End with a short "Audit" section listing: anything you had to assume, any phrase a reviewer should double-check for bias, and any fact that needs human verification.

NEGATIVE CONSTRAINTS:
- Never score, rank, filter, or describe a person using protected characteristics (age, gender, race, religion, disability, family status, nationality) or proxies for them (accent, appearance, name, school prestige, "polish").
- Never invent job requirements, salary figures, benefits, or company policies.
- Never draft a final decision (hire / reject / terminate / promote). Drafts and structures only — the decision belongs to a human.

OUTPUT FORMAT: the completed task first, then the "Questions before this is final" list, then the "Audit" section.
Best model for this: Claude Sonnet 5.5 for drafting and language review; GPT-6.1 for structured, high-volume tasks.

Worked example — the master prompt in action

Filled inputs: Task: draft structured interview feedback summary. Role: Senior Product Designer. Facts supplied: interviewer's raw notes ("strong portfolio walkthrough, explained trade-offs on the checkout redesign, struggled to answer metrics question, gave two examples of mentoring juniors"). Audience: the hiring panel.

Filled prompt (excerpt): the context block carries those notes verbatim; the instructions forbid adding any observation not in the notes.

Output excerpt:

Summary: The candidate gave a detailed walkthrough of a checkout redesign, explaining the trade-offs behind two key decisions. They described two specific examples of mentoring junior designers. Their answer on product metrics was less developed and would benefit from a follow-up question in the next round.
Audit: No assumptions added. One phrase in the raw notes ("polished presenter") was excluded from the summary because it describes presentation style rather than job evidence.

That audit line is the point: the model caught and removed a subjective impression instead of laundering it into the official record.

Guardrails — the rules every page in this cluster follows

Frequently Asked Questions

Practical answers covering model choice, workflow integration, and compliance rules.

For the work on this page, Claude Sonnet 5.5 is the strongest default for drafting and language review — job descriptions, outreach, feedback summaries — because it follows long, constraint-heavy instructions reliably. GPT-6.1 is excellent for structured outputs like scorecards and Boolean logic, and Gemini 4 Argon suits very long inputs such as a full hiring-manager intake document. The prompts here are plain text, so they also run in Copilot and free tiers; the generator's per-task recommendation tells you which model each template was tuned for.

It depends where the data goes. The generators on this page process everything in your browser and send nothing to a server, so drafting with them exposes no candidate data. The risk begins when you paste into a third-party chatbot: never paste candidate names combined with salary history, performance ratings, medical or disability information, protected characteristics, or personal contact details unless your organisation has an approved enterprise agreement covering that tool. Anonymise first — role titles and facts, not identities.

The tools themselves are general-purpose; compliance is about how you deploy them. Recruitment AI is high-risk under the EU AI Act, which brings transparency duties (telling candidates AI is involved), human-oversight requirements, and — for many deployers — a fundamental rights impact assessment under Article 27. Using a prompt to draft a job description is a long way from automated screening, but the moment AI output substantially influences who progresses, you should treat the deployment as high-risk and take legal advice.

Yes. Every template on this page is plain text and works in free tiers of ChatGPT, Claude, and Gemini. Free tiers have shorter context windows and may not hold a very long prompt plus a full intake document at once — if output quality drops, split the task: run the draft first, then run the audit pass as a second prompt. Nothing here requires a paid plan, an API key, or an account with PromptStaging.

Supply facts, not adjectives. "Senior engineer, fintech, must have led a migration from a monolith to services, team of four" produces a usable job description; "great senior engineer" produces a generic one. Then constrain the model: tell it what it may not add, what format to return, and make it list its assumptions at the end. Finally, iterate in small steps — change one input at a time and compare, exactly as you would with a Boolean string.

The anatomy of every prompt on this hub

Open any template in this cluster and you will find the same five bones. They are worth learning once, because they are what separate a prompt that survives contact with a real hiring process from one that produces a pleasant first draft and a problematic second one:

  1. Role. The model is told who it is — a recruitment copywriter, a technical sourcer, an assessment specialist. Role is not decoration: it changes which considerations the model prioritises, and each generator picks the specialist whose judgement the task actually needs.
  2. Context with brackets. Every variable fact sits in a [bracket]. Brackets are a contract: what is inside them is yours and true; everything else, the model must not invent. When you adapt a template for your own stack — a Custom GPT, a Claude Project, a Gemini Gem — keep the brackets and replace their contents per use.
  3. Numbered instructions. Steps, in order, with the output's structure specified down to headings and word limits. Vague format requests produce vague formats; a template that says "450 words maximum, these seven headings, in this order" gets exactly that.
  4. Negative constraints. The section most prompt lists skip entirely, and the one that matters most in HR. Every template here states what the model must never do: add requirements, invent facts, score protected characteristics, draft the final decision. In hiring, the failure modes live in this section.
  5. An audit step. The output ends by reporting on itself — assumptions made, phrases flagged, questions outstanding. Build the habit of reading the audit before the draft. It is the cheapest quality gate in your whole process.

If you write your own HR prompts from today onward, steal this skeleton whole. A one-line prompt becomes a governed workflow the moment it carries all five parts — role, bracketed context, numbered instructions, negative constraints, audit — and the difference shows up exactly where HR can least afford sloppiness: in the requirements nobody approved, the phrase nobody noticed, and the score nobody can explain.

Using these generators with the tools you already have

The generated prompts are plain text and deliberately tool-agnostic. Run them in the chat interface of GPT-6.1 / GPT-4o, Claude Sonnet 5.5, or Gemini 4 Argon; save a filled template as a Custom GPT, a Claude Project instruction, or a Gemini Gem if your team repeats the same task weekly; or keep the master prompt from this page pinned in your team wiki and specialise it by hand. If your organisation has an approved enterprise AI agreement, run the prompts inside it whenever candidate-adjacent material is involved — the in-browser generators here never transmit your inputs, but the moment you paste a generated prompt plus real candidate facts into a third-party tool, your organisation's data rules are the ones that govern. When in doubt: role titles and process facts travel; names, contact details, salary histories, and interview notes about identifiable people do not.

Internal links

---