Blog / 7 min read · 2026-08-16

Gemini Prompts for HR: Job Descriptions, Interview Kits, and Difficult Conversations

Use a Gemini prompt for HR to draft job descriptions, build structured interview kits, prepare policy communications, and plan difficult workplace conversations.

A good Gemini prompt for HR turns a role brief, a policy change, or a looming difficult conversation into a structured draft you can shape quickly. Used carelessly, the same tool amplifies the two classic HR failure modes: biased language that quietly narrows your candidate pool, and generic process text that ignores your actual policy and local employment law. The prompts below are built around both risks.

Ground rules first. No employee-identifiable information goes into the tool — no names, no performance details tied to a person, no health or grievance specifics. Describe situations generically or use placeholders. And employment law stays with counsel and your own policies: Gemini does not know your jurisdiction’s current rules, and HR decisions carrying legal consequence — terminations, accommodations, investigations — follow your process, not a chatbot draft.

How to write a useful HR prompt

Name the audience, the jurisdiction-sensitivity (“flag anything that may depend on local law” beats trusting the model to know it), the tone, and the format. Then ask for flags, not just prose — bias flags, assumption flags, legal-dependency flags.

Avoid a vague request such as: “Write a job description for a marketing manager.”

Use a request like this:

Draft a job description from the role brief below. Separate must-have requirements from nice-to-haves — maximum 5 must-haves, each one genuinely disqualifying if absent. Flag any wording that could discourage qualified candidates (gendered phrasing, age-coded terms like "digital native," unnecessary degree requirements), listing the flagged phrase and a neutral alternative. Do not add requirements I did not state.

Habits that improve almost every result:

• Separate must-have from nice-to-have explicitly — inflated requirements lists are the top JD failure.
• Ask for a bias pass as a labelled output section, every time.
• Mark anything jurisdiction-dependent “check locally” rather than letting the model guess at law.
• Keep every scenario generic; personalize afterwards, outside the tool.
• Treat outputs as drafts for HR judgment and, where it matters, legal review.

Prompt for job descriptions and hiring

Copy-paste prompt: job description with bias pass

Draft a job description from this brief.

Role: [title and team]
What this person will actually do (from the hiring manager): [paste]
Must-have requirements (I will cap these): [list]
Nice-to-haves: [list]
What we offer (factual, no superlatives): [paste]

Rules:
- Responsibilities as outcomes ("own monthly reporting end-to-end"), not activity lists.
- Keep my must-have/nice-to-have split; if a must-have looks like a proxy (degree standing in for a skill), flag it and propose the skill instead.
- Bias pass as a separate section: flagged phrase → why → neutral alternative.
- Plain language, no "rockstar/ninja/family," under 450 words.
- End with 3 screening questions a recruiter could ask that map directly to the must-haves.

Copy-paste prompt: structured interview kit

Build a structured interview kit for the role below. Structured means: same questions, same order, anchored scoring — I want consistency across candidates.

Role: [title]
Competencies to assess (max 5): [list]
Interview length: [minutes]

For each competency:
- One primary behavioral question ("tell me about a time…") targeting it
- One follow-up probe for thin answers
- Scoring anchors: what a 1, 3, and 5 answer concretely sounds like — anchors must describe the answer's content, not the candidate's confidence
- One red flag that sounds impressive but isn't (rehearsed war story with no personal role, blaming everyone else, results with no method)

Add timing per section summing to the interview length, and a note on what this kit deliberately does not assess.

Prompt for policy drafting and communication

Gemini drafts policy communication well; it must not originate policy content. The content comes from your decisions and counsel — the model’s job is making it readable and surfacing the questions employees will actually ask.

Copy-paste prompt: policy announcement employees will read

Draft an announcement of the policy change below for all staff.

What is changing (decided and approved): [paste]
Effective date: [date]
Why (the real reason, as we are willing to state it): [paste]
What employees need to do: [paste]

Rules:
- Lead with what changes for the reader, not with corporate framing.
- "What this means for you" section with concrete before/after examples.
- Anticipated-questions section: generate the 5 questions employees will most likely ask, with answers drawn only from what I provided — where my input does not answer one, write "to be addressed" so I see the gap.
- No policy-speak ("pursuant to," "be advised"). Under 400 words.
- Flag anything in my input that reads as legally sensitive with "review with counsel."

Copy-paste prompt: policy plain-language pass

Rewrite the policy extract below in plain language for the employee handbook, keeping the rules identical.

[paste extract]

Rules:
- Do not add, soften, or remove any requirement — meaning-preserving rewrite only.
- Second person ("you"), active voice, one rule per paragraph.
- Return the rewrite, then a table: original phrase → rewritten phrase → anything a reviewer should double-check because simplification could have shifted meaning.
- List terms that remain necessarily technical and give a one-line definition for each.

Prompt for difficult conversations and performance

The writing that precedes a hard conversation — the structure, the examples, the phrasing you will not fumble under stress — is legitimate prep work. The judgment about the person stays human, and the specifics stay out of the tool.

Copy-paste prompt: conversation plan (generic scenario)

Help me plan a difficult workplace conversation. The scenario below is generic — no real names or identifying details.

Scenario type: [e.g. repeated missed deadlines by an otherwise strong performer / peer conflict affecting a team / declining a promotion request]
What I can observe factually: [paste generic observations]
Outcome I need from the conversation: [paste]

Return:
1. An opening that names the topic in the first two sentences without ambush framing
2. The 3 factual observations to anchor on, phrased as behavior-and-impact ("when X happens, the effect is Y") — strip any wording that assigns motive
3. Questions that give the other person genuine room to respond, not rhetorical setups
4. The commitment structure to land on: what changes, how we will both know, when we check in
5. Phrases to avoid in this scenario type and what to say instead

Do not script their side. Where the right move depends on what they say, present it as a branch.

Copy-paste prompt: performance feedback de-vaguing

Below are draft feedback notes. Rewrite each as behavior-and-impact feedback, and flag what I still owe.

[paste notes like "communication needs work," "not enough ownership"]

For each note:
- If it names observable behavior: tighten to behavior → impact → expectation.
- If it is a trait judgment ("lacks initiative"): mark [NEEDS EVIDENCE] and list what specific observations I must supply before this is usable feedback.
- Strip mind-reading ("doesn't care about quality") and replace with the observable gap.

End with a count: usable now vs needs evidence. Feedback with no example is not deliverable.

Human review boundaries for Gemini in HR

Use Gemini for drafting, structuring, bias passes, and conversation prep. Keep on the human side:

• Employee privacy. No names, no performance or health specifics tied to an identifiable person, no grievance details. Generic scenarios only.
• Decisions about people. Hiring, rating, promotion, discipline, termination — a model shortlisting or rating candidates is both bad practice and, in a growing number of jurisdictions, regulated automated employment decision-making. Drafts support the decision-maker; they are never the decision.
• Employment law. Notice periods, protected characteristics, accommodation duties, investigation procedure — jurisdiction-specific and consequence-bearing. “Check locally” flags in these prompts exist because the model genuinely does not know.
• Investigations and grievances. Follow your documented procedure exactly; AI drafting in a live investigation creates discoverable text you did not control.
• Anything that ends up in an employee’s file gets human authorship and review.

A useful internal rule: Gemini prepares you for the human parts; it never replaces them.

FAQ

Can HR use Gemini to screen resumes?

Ranking or filtering candidates with AI is regulated in a growing number of places (automated employment decision rules) and risky everywhere — bias in, bias out, at scale. Use it instead to build better screening criteria and structured interview kits, and keep candidate evaluation with humans.

Can Gemini write job descriptions?

Yes, well — if you cap must-haves, feed it the real role content, and demand a bias pass as a labelled section. The classic failures (requirement inflation, coded language, activity lists instead of outcomes) are exactly what the prompt has to forbid.

Is it safe to discuss employee issues with Gemini?

Only as generic scenarios: no names, no identifying details, no specifics from a live case. Conversation structure and phrasing prep work fine generically; a live grievance or investigation does not belong in a third-party tool at all.

Can Gemini write HR policies?

It can turn decided policy into readable communication and plain-language handbook text. The policy content itself — what the rules are — comes from your decisions and counsel, because the model does not know your jurisdiction’s requirements and will not be accountable for getting them wrong.

Can Gemini help with performance reviews?

As a de-vaguing tool for your own draft feedback — turning trait judgments into behavior-and-impact statements and flagging claims that lack evidence. The assessment of the person is yours; text about identifiable employees stays out of the tool.

Conclusion

The most useful Gemini prompt for HR states the audience, demands flags — bias, assumptions, jurisdiction — and keeps every scenario generic. Start with the job-description, interview-kit, policy-communication, and conversation-plan prompts above, and let the decisions about actual people stay where they must: with people.

More profession-specific sets: Gemini prompts for lawyers, Gemini prompts for teachers, or the prompt gallery.