Overview

You've been reading job descriptions for 15 years. You can spot the template from a mile away: "Dynamic organization seeks talented professional for fast-paced role. Requires 10+ years of experience. Responsibilities include managing projects, collaborating with stakeholders, and driving results."

Three candidates apply. All overqualified and unfocused. None of them are who you're actually looking for.

The problem isn't the format. It's that your job description sounds like every other job description. It doesn't tell a real person why they'd want to work for you. It doesn't describe the actual role, the actual challenges, the actual team. It's a template with a job title plugged in.

This lesson teaches you to use AI to write job descriptions that are specific, compelling, and honest. You'll learn how to avoid the bias-laden language that AI sometimes generates. You'll see how to optimize for the right search and the right candidates. And you'll learn the difference between a JD that gets 50 applications and a JD that gets 5, but the 5 are actually qualified.

Why This Matters for HR Professionals

Job descriptions are your first marketing message to candidates. They're also a legal document that defines the role, the requirements, and, implicitly, what's expected. If they're bad, everything downstream gets harder.

Bad JDs attract:
- Over-qualified people who get bored and leave
- Under-qualified people who you have to screen and reject
- People who don't fit the culture but match the template
- Fewer diverse candidates (because the language is coded toward a specific demographic)

Good JDs attract:
- People who actually want to do the job
- People who fit the role, the level, and the culture
- Diverse candidate pools (because the language is inclusive)
- People who self-select in (because the JD is honest about challenges) and self-select out (because they know it's not for them)

AI can help you move from generic templates to specific, compelling job descriptions. But you need to know how to guide it, and you need to catch the bias and accuracy problems AI creates.

Anatomy of a Strong Job Description

Before we use AI, let's be absolutely clear on what makes a job description work. A strong JD does three things:
1. It attracts qualified people who actually want the job
2. It sets clear expectations so people self-select in (or out)
3. It legally protects you by defining the role and requirements

A strong JD has these essential components:

• 
  Role title and level (Senior X, not just X)
  - This matters for search engines and for candidate expectations
  - "Marketing Manager" is different from "Senior Marketing Manager" from "Marketing Manager, Demand Gen"
• 
  A compelling opening (why someone would care about THIS role, not just any role)
  - Don't start with a mission statement
  - Start with: What's interesting about this specific role? What problem are you solving? Why would someone want to do this?
• 
  About the company (who you are, what you actually do, what your culture is really like)
  - Not the website version
  - The real version: "We're a 120-person B2B SaaS company. We move fast. We're profitable, not VC-backed. Our customers are logistics companies."
• 
  The role in context (who does this person report to, who's on the team, what's the scope)
  - "You'll be the second marketer on a team with our VP of Marketing"
  - Helps candidates understand what they're joining
• 
  Key responsibilities (4-6 actual responsibilities, not a laundry list)
  - What will this person spend 80% of their time on?
  - Not everything they might do, just the big things
• 
  Required qualifications (experience, skills, knowledge, be realistic, avoid inflated requirements)
  - Be honest: Is 10 years really required, or would 3 strong years work?
  - Be specific: What does competence actually look like?
• 
  Nice-to-haves (these don't appear in many JDs, but they help candidates self-assess)
  - "It's great if you have X, but not required"
  - Helps diverse candidates self-select in (rather than self-selecting out because they don't match perfectly)
• 
  Why someone would love this role (career development, impact, team, culture)
  - What's in it for them?
  - Be honest: Is this a stepping stone? A stable job? A growth opportunity?
• 
  Practical logistics (salary range, location, work setup)
  - Candidates want to know. Transparency attracts serious candidates.

That's the structure. Now let's use AI to draft it well.

The AI-Assisted JD Process

Step 1: Gather Your Actual Requirements

Before you ask AI to draft anything, you need to know what the role actually is. Most HR people skip this step. They ask AI to write a JD for "a marketing manager" and expect magic.

You need:
- Who will this person report to? (Title and a sentence about that person)
- What team are they joining? (How many people? What's the team's current state?)
- What are the actual top 3-4 things they'll spend their time on?
- What will success look like after 6 months?
- What are the obstacles or challenges in this role?
- What's your company like to work for? (Culture, growth, challenges, pace)

Write this out in a paragraph or two. This becomes context for your AI prompt.

Example (real context, not generic):
"Hiring a Marketing Manager for our 120-person B2B SaaS company. She'll report to our VP of Marketing. The marketing team is currently the VP and one senior coordinator; this is the second marketer. We're scaling from 'marketing is happening' to 'marketing has a real strategy.' The main focus: content marketing and thought leadership (building authority in our vertical), demand gen campaigns, and supporting sales enablement. Not graphic design or video. We outsource that. Success at 6 months: people know who we are in our space, lead flow is up 30%, sales team has materials they actually use. Challenges: we're early, so there's ambiguity; things change; the VP is strategic but not operational. Culture: direct, collaborative, everyone contributes ideas, we move fast."

When you write this down, you understand the role better. Now you prompt AI with this context.

Step 2: The AI Prompt

Using your gathered context:

"I need a job description for a Marketing Manager role. Here's the context: [paste what you wrote above]. The role level is Mid/Senior. We want someone who understands that content and thought leadership are important, who can manage ambiguity and move fast, and who wants to build something (not just execute someone else's plan). Tone: warm and direct, not corporate. Honest about challenges but also about why this matters. Avoid: corporate jargon like 'fast-paced' or 'strategic partner.' Don't inflate requirements. We'd rather have 3 years of strong experience than 10 years of inflated titles. Length: 500-600 words. Format: Compelling opening (2-3 sentences explaining why someone would care), company overview (2-3 sentences), the role (1 paragraph explaining what they'll actually do), responsibilities (4-5 bullet points), requirements (be realistic), nice-to-haves, why you'll love working here (specific, honest), compensation and logistics (we'll fill this in)."

This prompt is specific. It gives AI the actual context, the tone you want, what to avoid, and the structure.

Step 3: Review and Refine

When AI returns the draft, you're not done. You're checking:

• Is the role description accurate? Does it actually describe what this person will do?
  - Is the tone right? Does it sound like your company or like a template?
  - Are the requirements realistic? Would you actually reject someone who didn't meet these?
  - Is there bias language? Search for "young," "energetic," "native," "fit," or gendered language.
  - Is anything missing? What would a candidate need to know?

Then you iterate. You tell AI what to adjust: "Good start. The opening is too formal. Make it more personal. Cut the 'why we're great' section in half. Add something about team dynamics. Change 'must have 5+ years' to 'prefer 3+ years of'. We're being too strict."

Most AI-drafted JDs need 1-2 rounds of refinement. That's normal.

Bias in Job Descriptions: The Silent Killer

This deserves its own section because it's where JDs fail most often, and AI makes it worse.

Bias types in JDs:

• Age bias: "Young," "energetic," "digital native," "startup experience," "recent grad energy," "keep up with the pace"
  - Gender bias: Gendered words applied differently ("aggressive" vs "assertive," "bossy" vs "decisive")
  - Ability bias: "Must have stamina," "high-energy environment," "fast-paced," "always available"
  - Cultural bias: "Culture fit" (code for "like us"), "love working hard," "willing to go above and beyond"
  - Credential inflation: "10+ years required" when 5 years would actually work
  - Coded language: "Passionate," "hungry," "driven," "will move mountains"

AI picks up these biases from training data and reproduces them. You have to catch and remove them.

Scan your AI-drafted JD for these patterns:
- Search for "young," "energetic," "native," "stamina," "fit," "passionate," "aggressive"
- Check if you're using gender-coded words
- Look for inflated "must have" requirements
- Challenge any "always," "requires," or "must be" language that's actually a nice-to-have

Example of bias language:
"We're seeking a young, energetic Marketing Manager who is passionate about building startups. Must have 10+ years of enterprise software marketing and be willing to work long hours to drive results. You'll need to be aggressive in your approach and excited to take on any task. Culture fit is essential."

Problems:
- "Young, energetic" = age discrimination
- "Passionate about building startups" = assumes personal life choices (no parents, no caregivers)
- "10+ years" = probably too strict, and is age coding
- "Long hours," "aggressive," "any task" = ability bias, work-life balance concerns
- "Culture fit" = code for homogeneous hiring

Revised:
"We're seeking a Marketing Manager with 3+ years of B2B software marketing experience. You understand how to build authority in a space through content and thought leadership. You're comfortable with ambiguity and can move fast. You're collaborative and direct. You're interested in helping us scale marketing from 'happening' to 'strategic.' This is a mid-level role with real influence on company direction."

Better because:
- Realistic experience requirement
- Specific about what you actually need
- Removed bias language
- Honest about growth/pace
- No jargon

Important: Always scan your JD against the bias checklist. Age, gender, ability, culture fit language will narrow your candidate pool and might expose you legally.

Optimizing for Search and Discoverability

Job descriptions don't just go to people reading them. They also get indexed and searched on job boards. If you write them in a way that's not searchable, qualified candidates won't find you.

This is where "job title" matters more than you think. If you're hiring a "Marketing Manager" but your internal title is "Director of Demand Generation," you're not reaching people searching for your job title. You need to use language candidates actually search for.

How AI helps here:
- Ask: "What would someone search for when looking for a role like this?"
- AI will suggest: "People searching for marketing manager, content marketer, demand gen, growth marketing, or marketing coordinator might be interested. How do you position this?"
- This helps you title and describe the role in searchable language.

In your JD, include:
- The job title people actually search for
- Keywords for the role (not keyword-stuffed, just naturally included)
- Level language ("Senior," "Mid-level," "Entry-level") people filter by
- Industry language ("SaaS," "B2B," whatever applies)

Example:
Instead of: "Join our organization in a marketing role"
Try: "Marketing Manager (Content Marketing, B2B SaaS)": because that's what people search for

Common JD Mistakes That AI Can Amplify

When using AI to draft job descriptions, watch out for these mistakes. AI will happily repeat them if you don't catch them:

Mistake 1: Inflated requirements
Bad: "10+ years of X required"
Better: "3+ years of X with demonstrated mastery"

AI will copy the pattern from existing JDs that have inflated requirements. You need to sanity-check every requirement: "Would we really reject someone with 5 years of strong experience?"

Mistake 2: Unspecific responsibilities
Bad: "Drive strategic initiatives and optimize performance metrics"
Better: "Build demand generation campaigns, measure effectiveness, and report quarterly results to leadership"

AI can default to vague corporate language. Push it toward specific, observable work.

Mistake 3: Outdated language
Bad: "Fast-paced startup"
Better: "We ship quickly but thoughtfully"

Outdated phrases make your JD sound generic. Tell AI: "We're not a 'fast-paced startup' anymore. Describe what our pace and culture actually like."

Mistake 4: Bias language
Bad: "Looking for a hungry, energetic team player"
Better: "We're looking for someone who's motivated, collaborative, and interested in growth"

AI picks up biased language from training data. Always scan for age bias (young, energetic, native), gender bias (aggressive vs. assertive), and cultural bias (team player, work-life balance language that codes for life stage/caregiving status).

Practical Workflow: From Context to Published JD

Here's the full workflow that works:

Day 1: Gather Context
- Spend 20 minutes with the hiring manager
- Ask: Who reports to whom? What's the team? What does success look like? What's hard about this role? What's the real culture?
- Write notes in plain language
- You now have your context

Day 2: Prompt AI
- Use your context as the prompt
- Ask AI to draft the JD with specific tone and structure
- AI generates a draft
- You spend 30 minutes reviewing and marking up issues

Day 3: Refine
- Ask AI to revise: "Make it less corporate. Be more specific about the role. Remove the jargon."
- Compare original to revision
- Pick the best parts of each
- Do a bias scan using the checklist

Day 4: Finalize
- Read it out loud
- Does it sound like your company?
- Add specific details: salary range, real location, real reporting structure
- Get hiring manager approval
- Legal review if it's new to your company
- Publish

Ongoing: Reuse and Adapt
- You now have a template for this role type
- Next time you hire this role, you start with this JD
- Adapt for specific context
- Reuse 80%, customize 20%

This workflow takes 4 days and saves you hours down the road through reuse.

Try This Now: Three Exercises

Exercise 1: Gather Your Context

Pick a role you're actually hiring for (or a recent one you filled). Write down:
- Who reports to whom
- What's the actual team situation
- What 3-4 things will they spend their time on
- What's success look like at 6 months
- What's the challenge in this role
- What's your company culture actually like (not the website version, the real version)

This is your context. This is what you'll feed to AI.

Exercise 2: Draft, Scan for Bias, Refine

• Ask AI to draft a JD using your context from Exercise 1.
  - Scan it for the bias language list above. Mark any problematic language.
  - Ask AI to remove bias language: "Remove any age, gender, ability, or culture-fit language. Be realistic about requirements."
  - Compare old vs new. You now have a bias-free JD.

Exercise 3: Test for Tone and Specificity

Paste your final AI-drafted JD into a document. Read it out loud. Does it sound like your company or like a template?

If it sounds like a template, ask AI: "Make this more specific to our company. Don't just say we're collaborative, describe what that looks like. Give examples of the work. Mention specific challenges."

Then compare. Notice what changed. That's how you learn what AI needs to produce your voice.

Practical Application - "What to Do Monday Morning"

• 
  Create a JD context template. Save a document with prompts: "Who does this person report to? What's the actual team? What will they spend 80% of their time on?" Use this every time you hire.
• 
  Before you ask AI for a JD, write your context. Spend 15 minutes capturing what you know. This makes AI output infinitely better.
• 
  Create a bias checklist. Print or bookmark this lesson's bias list. Scan every JD against it before posting.
• 
  Iterate once or twice. First AI output is rarely perfect. Tell AI what to adjust. Get to something you're happy with.
• 
  Verify details. Compensation range, location setup, reporting structure. These should be accurate. Double-check them before posting.

Key Takeaways

• Gather actual context before asking AI: The more specific your prompt, the better the output.
  - Structure matters: Opening, company context, role description, responsibilities, requirements, nice-to-haves, culture fit.
  - Scan for bias: Age, gender, ability, culture fit language will narrow your pool and create legal risk.
  - Be realistic about requirements: "10+ years" is almost always inflated. What do you actually need?
  - Optimize for search: Use language candidates search for. Title it searchably.
  - Iterate: First draft is a starting point. Refine for tone, specificity, and accuracy.

FAQ

Q: How do I avoid the "culture fit" trap?
A: Instead of "culture fit," describe the actual culture. "We're direct and challenge each other." "We value thoughtfulness and deliberate decisions." These are specific behaviors, not personality types.

Q: Should I include salary range?
A: Yes. Candidates want to know. AI can help you think through what's competitive. But you should know your range before asking AI.

Q: What if my actual role doesn't match what's on the job board?
A: That's common. Pick the title candidates search for, not your internal title. If internally you call it "Demand Gen Manager" but candidates search "Marketing Manager," use the searchable title.

Q: How do I know if my requirements are realistic?
A: Ask yourself: "Would I reject someone who didn't have this?" If the answer is no, remove it. "Prefer" is honesty; "require" should be non-negotiable.

Q: Can I use the same JD if I'm rehiring for a role someone left?
A: Update it. Things have changed. Ask the hiring manager: "What worked about the last person? What would be different about an ideal next person?" That becomes your refresh.

What's Next

A great job description gets the right candidates to apply. Lesson 2.2 is about what comes next: how to craft outreach messages to candidates you find (not just those who apply). You'll learn to write recruiting emails that feel personal, not generic, and that actually get responses.