More requirements doesn’t mean a better candidate. A fifteen-item “required skills” list isn’t a quality filter — it’s a filter for who’s willing to apply despite not meeting half the bar, which skews toward overconfident candidates rather than qualified ones.

If you’ve never written a job description before, or you’re stuck reusing a template from 2019 that nobody’s looked at critically since, you’re not alone. Most hiring managers learn job description writing by copying whatever the last person used — which means outdated structure, bloated requirements, and language nobody has ever actually checked for bias get passed down indefinitely.

This guide gives you how to write a job description using AI as a complete, repeatable system: a six-section structure that covers everything a JD needs, the exact prompts to generate each section, a dedicated bias-audit step most guides skip entirely, and a fix for the single most common JD mistake — requirement bloat. You’ll also get a compliance checklist and a way to reformat one JD for your ATS, LinkedIn, and Indeed without rewriting it three times.

One thing worth knowing upfront: this article is about building the internal document itself — the structurally complete, compliance-aware job description your hiring manager is accountable for. If you’re specifically trying to reduce unqualified applicant volume on a public job posting, our companion guide on writing a job posting using AI that filters right covers the Outcomes-Based Filter Framework for that specific problem — the two work well together.

Why Most AI-Written Job Descriptions Sound the Same (And Attract the Wrong Candidates)

Here’s what actually matters: a one-line prompt like “write a job description for a Marketing Manager” gives the AI nothing to differentiate your role from the thousand other Marketing Manager postings in its training data. So it defaults to the safest, most generic version — “fast-paced environment,” “team player,” “5+ years experience” — language that’s technically accurate but tells a candidate nothing real about the job.

A common mistake is assuming the fix is a better AI model. In practice, the model matters far less than the structure you give it. The same vague prompt produces the same generic output across ChatGPT, Copilot, and Gemini, because none of them have been told what a complete, well-built job description actually contains.

What many people overlook is that this isn’t a minor structural nitpick — a job description missing one of these six sections (most commonly the compliance statement, or a requirements list that was never capped) is a document that either repels qualified candidates or creates downstream legal exposure. The fix is giving AI the full structure up front, not patching a generic draft after the fact.

The 6-Section Job Description Framework

In practice, every well-built job description breaks down into these six sections. Skip one, and the document either fails to inform candidates properly or leaves your company exposed on a compliance question nobody thought to ask.

Role Summary

A common mistake is opening with the company’s mission statement instead of the role itself. The role summary should answer one question in two to three sentences: what does this person actually do, and why does the position exist? Candidates skim this first — if it’s vague, they move on before reaching the responsibilities section.

Key Responsibilities

This becomes important when the list runs too long or too short. Five to eight specific, action-oriented duties is the sweet spot — enough to show real scope, not so many that the role reads as an impossible catch-all. Each line should start with a verb: “lead,” “manage,” “build,” not “responsible for.”

Required vs. Preferred Qualifications (The 6-Item Rule)

The reality is this is where most job descriptions go wrong, and we cover the full fix in the requirement bloat section below. The short version: cap required qualifications at six genuinely essential items, and move everything else — anything trainable within 90 days — into a separate preferred list.

What We Offer

This section needs compensation range (required by law in a growing number of states as of 2026), core benefits, and one or two genuine culture details — not generic phrases like “great culture” that say nothing. Specificity here does real work: “fully remote, async-friendly” tells a candidate something concrete that “flexible work environment” doesn’t.

Equal Opportunity & Compliance Statement

What many people overlook is that AI won’t add this section unless you explicitly ask for it — and a generic AI prompt has no way of knowing your state’s specific disclosure requirements. We cover exactly what to check for in the compliance review section near the end of this guide.

The Complete Copy-Paste Prompt to Write a Job Description Using AI

In practice, this single prompt builds all six sections in one pass. Fill in the five bullet points about the role, and run it as-is — no prompting expertise required.

Act as an HR content specialist writing a job description for [ROLE TITLE] at [COMPANY NAME / TYPE].

Using the 5 bullet points below, write a complete job description with exactly these six sections, in this order:

1. ROLE SUMMARY (2–3 sentences) — what this person actually does and why the role exists.
2. KEY RESPONSIBILITIES (5–8 bullet points) — each starting with an action verb.
3. REQUIRED QUALIFICATIONS (maximum 6 items) — only skills genuinely needed to perform the role on day one.
4. PREFERRED QUALIFICATIONS (bullet points) — trainable or nice-to-have skills, no cap.
5. WHAT WE OFFER (bullet points) — compensation range, benefits, 1–2 specific culture details.
6. EQUAL OPPORTUNITY STATEMENT — standard EEO language.

Do not use "rockstar," "ninja," "ninja-level," "guru," or "fast-paced environment." Keep tone professional and specific, not generic.

Role details:
[PASTE 5 BULLET POINTS: role/team, 3-4 key tasks, must-have skills, location/remote status, salary range if known]

The reality is that getting consistent output isn’t about finding clever phrasing — it’s about giving the AI a fixed six-section structure and an explicit cap on requirements before it starts writing, not after.

How to Audit a Job Description for Bias Using AI

Bias in job descriptions is almost never intentional — it’s inherited. Someone wrote “must thrive in a fast-paced, high-pressure environment” years ago, and it’s been copy-pasted into every posting since, quietly signaling something to candidates who’d otherwise be a great fit. AI is genuinely useful here, not because it’s neutral by default, but because it can be explicitly instructed to flag patterns a busy hiring manager skims right past.

The reality is this audit step is what most competing guides treat as an afterthought — a single line like “make sure it’s inclusive” with no actual process behind it. Here’s the process:

Review this job description for gender-coded, age-coded, or otherwise exclusionary language.

For each flagged word or phrase, explain why it may discourage certain applicants and suggest a neutral replacement. Do not rewrite the whole document — return a table with three columns: Original Phrase, Why It's a Concern, Suggested Replacement.

Job description:
[PASTE JD]

This becomes important when you realize what the table format actually does: it gives you a defensible, reviewable paper trail instead of a silent rewrite you’d have to trust blindly. You can see exactly what was flagged and why, and accept or reject each suggestion individually.

For a deeper, more general technique you can apply beyond job descriptions specifically, our guide on why AI sometimes gets things wrong explains the underlying reason audits like this matter — AI reflects patterns in its training data, including biased ones, unless explicitly instructed otherwise.

Fixing Requirement Bloat: The 6-Skill Rule

Every hiring manager thinks more requirements means a better candidate. It’s the opposite. A fifteen-item required list isn’t a filter for quality — it’s a filter for who’s willing to apply despite not meeting half the bar, which skews heavily toward overconfident candidates rather than qualified ones. Research consistently shows this discourages otherwise-qualified candidates, particularly women, from applying at all, since many people only apply when they meet nearly every listed requirement.

The fix isn’t deleting skills the hiring manager genuinely wants — it’s separating “needed on day one” from “nice to have, trainable within 90 days,” and being honest about which is which.

Here is a list of skills a hiring manager wants in a candidate. Split them into two categories:

REQUIRED (maximum 6 — only skills needed to perform the role on day one)
PREFERRED (everything else, including anything trainable within 90 days)

Briefly justify each REQUIRED item in one sentence.

Skills list:
[PASTE FULL SKILLS LIST]

What many people overlook is that this prompt does something a recruiter’s manual negotiation rarely accomplishes cleanly — it gives the hiring manager a reasoned justification for each cut, rather than feeling like their wishlist is being arbitrarily trimmed. A list with explained reasoning is far easier to agree to than a list with items silently removed.

Repurposing One Job Description for LinkedIn, Indeed, and Your ATS

A job description isn’t one document — it’s three. What your ATS needs for searchability, what’s legally required in the formal record, and what a candidate scrolling LinkedIn on their phone actually reads are three different jobs for the same content. Companies that write one version and paste it everywhere are leaving applicants on the table, because a dense, formal JD that works fine in an ATS reads like a wall of text on LinkedIn.

Using this master job description, create three versions:

1. FORMAL ATS VERSION — unchanged structure, full detail, all six sections intact.
2. LINKEDIN VERSION — conversational tone, under 200 words, with 3 bullet highlights instead of the full responsibilities list.
3. INDEED VERSION — under 150 words, keyword-dense for search, front-load the role title and key qualifications.

Job description:
[PASTE MASTER JD]

In practice, this single batch request replaces what used to be three separate rewriting passes. Recruiting coordinators and agency recruiters posting the same role across multiple platforms see the biggest time savings here — what was a 25–30 minute manual reformatting task becomes a single prompt.

ChatGPT vs. Copilot vs. Gemini: Which Is Best for Writing Job Descriptions?

The reality is each tool fits a different part of this workflow, and most HR teams end up using more than one depending on where their hiring documents already live.

ChatGPT gives the most flexible prompting and lets you save the master prompt as a reusable Custom Instruction, so every new chat already knows your six-section structure and banned-word list without retyping it. Microsoft Copilot, when used inside Word, lets HR teams draft directly in the company’s existing template — useful if your hiring managers already live in Word and Outlook, since it removes a manual copy-paste cleanup step. See drafting job descriptions directly inside Microsoft Word with Copilot for the native workflow. Google Gemini supports reusable Gems for teams that want a saved, repeatable JD template inside Google Workspace — see building a reusable job description template with Gemini.

What to Always Review Before Publishing an AI-Written Job Description

This is review guidance, not legal advice — for anything jurisdiction-specific, confirm with your legal or compliance team. But there are five things worth checking on every AI-written job description before it goes live, every time, regardless of how confident you feel about the draft.

Review this draft job description as an HR compliance checklist (not legal advice). Flag whether it is missing:

- A salary range
- An EEO statement
- ADA-compliant physical requirement language (if applicable to the role)
- Any age- or gender-coded terms not already caught in a prior bias audit

List what's missing as bullet points, with a one-line note on why each matters.

Job description:
[PASTE JD]

What many people overlook is that this prompt surfaces gaps in under a minute that might otherwise go unnoticed until a candidate complaint or a compliance audit. It won’t replace legal review for genuinely complex situations, but it catches the common, avoidable gaps before they become a problem.

It is not safe to paste specific salary bands or internal compensation structures into the free, public version of ChatGPT without checking your settings, since that data may be used to train future models depending on your account’s data controls. For sensitive compensation details, use an enterprise-tier tool like Microsoft 365 Copilot or ChatGPT Team/Enterprise, where data is protected by commercial agreements and excluded from training. Review OpenAI’s enterprise data privacy standards directly, since exact settings and defaults can change.

Frequently Asked Questions

Next Steps