Skip to content
Home / How to Use AI to Write Better Job Descriptions

Writing

How to Use AI to Write Better Job Descriptions

Most AI-written job descriptions read like a wish list nobody could match. Here is how to use AI to write one a real person would actually want to apply to.

A recruiter reviews a candidate’s documents during a job interview in a modern office setting.
A recruiter reading a candidate’s documents — the moment your job description is supposed to lead to. Photograph: Tima Miroshnichenko / Pexels.

Most job descriptions read like a wish list nobody could match. Ten years of experience in a tool that has existed for four. A degree in a field the job barely touches. “Fast-paced environment” and “rockstar” and “wear many hats,” which is code for: this job is three jobs and the pay is for one.

AI made this worse before it made it better. Ask a chatbot to “write a job description for a marketing manager” and it will hand you the average of every job description ever written — which is exactly the problem. The average job description is a document designed to repel the people you actually want and attract the people who are good at applying to jobs.

I have written job descriptions with AI help for small teams that could not afford a recruiter, and the difference between the ones that worked and the ones that did not was never the AI. It was what I fed it, and what I refused to let it decide. Here is the method that works.

1. Most job descriptions are written to repel, not attract

Read a typical job description from the applicant’s side for once. The first paragraph is about the company — how innovative it is, how fast it is growing. Then comes a wall of requirements, then a shorter wall of “nice to haves,” then a vague line about benefits. Nowhere does it say what the person will actually do on a Tuesday.

That is backwards. A good candidate is not choosing between your job and unemployment; they are choosing between your job and three other jobs. The description is a sales document, and it is selling to the one person you most want to hire. If it reads like a compliance form, that person keeps scrolling.

Before I touch any AI tool, I ask one question: who is the person I am trying to reach, and what would make them stop scrolling? That answer — not a job title — is what the description is actually about.

2. Start with the job, not the job description

AI cannot write a good job description from a job title, because a job title is not a job. “Customer success manager” means ten different things at ten different companies. The AI will guess the most common one, and the most common one is wrong for you.

So I start by talking to the person who will manage this hire — or, if that is me, by writing down what the job actually is. What does this person do in their first week? What decisions do they make without asking anyone? What does a good month look like, in numbers? What is the hardest part of the job, honestly?

Fifteen minutes of real answers beats an hour of prompting. I write it as rough notes, the way I would explain the job to a friend over coffee: “They’ll own our onboarding emails, they’ll talk to about ten customers a week, and the hard part is that our product changes fast so the docs are always slightly out of date.” That paragraph is the raw material. Everything else is shaping.

3. Feed the AI the real job, not a job title

Here is the prompt that actually works: paste in those rough notes and ask the AI to turn them into a job description — then tell it what to leave out. “Write a job description from these notes. Keep it under 400 words. Use plain language. Do not add any requirement I did not mention. Do not use the words dynamic, passionate, ninja, rockstar, or fast-paced.”

The constraints matter more than the request. An AI with no constraints will add requirements you never asked for, because that is what job descriptions usually contain. It will invent a degree requirement. It will ask for five years of experience with a tool your team adopted last year. Every invented requirement is a real person deciding not to apply.

I also ask for two versions: one short and direct, one a little warmer. Then I take the best lines from each. AI drafts are raw material, same as always — the first version is never the one you publish.

4. Cut the requirements list in half — twice

This is the step that changes everything, and it is the step most people skip. Take the requirements list the AI produced and cut it in half. Then look at what is left and cut it in half again.

What survives is usually three or four things, and those are the actual job. Everything else was a wish, a habit, or a copy of someone else’s posting. “Bachelor’s degree required” — is it, really, or is that just what the last posting said? “Five years of experience” — doing what, exactly, and would three years plus the right project count? If you cannot defend a requirement in one sentence, it does not belong in the must-have list.

I split what remains into two honest lists: “You need” and “It helps if.” The first list is short and non-negotiable. The second list is everything else. Candidates read the first list and think, “I can do this.” They read the second list and think, “I know some of that.” That is the difference between an application and a scroll-past — and research has backed this up for years: long must-have lists discourage good candidates, especially the ones who do not overestimate themselves.

5. Write the parts only you can write

AI can structure a job description. It cannot tell a candidate why this job is actually good, because it does not know. That part is yours, and it is the part candidates actually read.

Three things belong in every job description, written by a human: what success looks like in 90 days, what is genuinely good about working here, and the salary range. “In your first 90 days you will have rebuilt our onboarding sequence and cut first-week support tickets by a third” tells a candidate more than five paragraphs of company mission. “We are a team of six, we work remotely, and nobody is expected to answer messages after 6pm” tells them more than “great culture.” And the salary range — say it. Hiding the salary does not save you money; it costs you the candidates who have better options, which is to say the candidates you want.

I write these three parts by hand, in plain language, before I let the AI near the draft. Then I ask the AI to fit the whole description around them — never the other way round.

6. The applicant test: read it as the person you want to hire

Before any job description goes live, I read it once as the person I want to hire. Not as the manager, not as the founder — as the experienced candidate with two other offers on the table.

That read catches what AI always gets wrong: the requirement that quietly excludes parents, or career-changers, or anyone without a particular degree. The phrase “recent graduate preferred” that is age discrimination wearing a friendly face. The “must be available at all hours” that tells a good candidate this job will eat their life. The jargon that only an insider would understand — if your description needs a glossary, the job will too, and that is not a selling point.

I also check the numbers by hand. If the description says “competitive salary,” I ask whether I would apply to a job that said that to me. The answer is usually no. Specificity is respect, and candidates can tell the difference between a company that respects their time and one that is just filling a template.

The honest truth

AI will not write you a good job description. It will write you the average job description, and the average job description is why good candidates ignore most postings. The job is the valuable part — the real work, the real team, the real trade-offs — and only you know what those are.

Use AI the way I do: as the fastest drafting assistant you have ever had, working from real notes about a real job. Real job first, short honest requirements second, human-written reasons to apply third, and a read-through as the applicant before it ships. That order is the whole trick. Get it right and your description stops sounding like a form. It starts sounding like an invitation — from a place that knows what it needs and respects the person reading it.

Frequently Asked Questions

Will an AI-written job description attract the wrong candidates?

It can, if you publish the first draft. The first draft is the average of every posting the AI has seen, and it will attract people who are good at matching generic requirements. The fix is in the inputs: feed it the real job in your own words, cut the requirements list hard, and read the result as the candidate you want. A description built that way filters for fit, not for keyword-matching.

Should I include the salary range in the job description?

Yes. Every time. Candidates with options skip postings without a range, and those are the candidates you want. A range also forces you to decide what the job is actually worth before you start interviewing, which saves everyone time. If your range is below market, that is useful information too — better to learn it from silence now than from a failed search in six weeks.

How do I stop AI from writing biased or exclusionary language?

Tell it explicitly, then check its work yourself. Ask for plain language, no degree requirement unless the job legally needs one, no “recent graduate,” “digital native,” “high-energy,” or other coded phrases, and no requirements you cannot defend in one sentence. Then read the draft as the applicant: if a line would make a capable person with a different background feel unwelcome, cut it. AI is good at following these rules once you state them; it is terrible at inventing them.

Can AI help rewrite an existing job description, or only write new ones?

It is often better at rewriting than at writing from scratch. Paste your current posting in and ask: “Cut this to half the length. Keep every real requirement. Remove anything that is a wish, a cliche, or copied from another company’s posting. Rewrite it in plain language for the person who will do the job.” Then add the human parts — the 90-day picture, the honest upsides, the salary range — yourself. Most old postings improve dramatically from that one pass.

Try it on your draft. Repetition finder runs in your browser. Paste a section of your job description in before you post it — it surfaces the words you lean on too often, so your posting doesn’t read like every other listing in the pile.

Continue your workflow

Was this guide helpful?

Your selection stays on this page. To send feedback to the publication, use our contact form.

About this publication

From Text Craft Pro

Practical guidance on clear writing, thoughtful editing, and useful web content. Our guides use illustrative examples and original diagrams, with AI-assisted drafting and editing.

Read our editorial approach About Text Craft Pro

Published 20 November 2026 · Text Craft Pro · AI-assisted educational writing. Editorial policy · Suggest a correction

Scroll to Top