Last updated: September 2026

Knowing how to write a job description that actually fills roles is a filtering problem disguised as a writing problem. Most companies do not lose good candidates because their ad is badly worded. They lose them because they post a 900-word list of duties, hide the salary, use an internal job title nobody searches for, and then wonder why the pipeline is empty three weeks later. AI changes the economics of fixing this. What used to cost an afternoon of staring at a blank template now takes twenty minutes of structured prompting, provided you feed the model real inputs: outcomes, metrics, team context, and a pay range. This guide walks through the complete workflow used by small teams across the US, UK, Europe, and Southeast Asia: a seven-step drafting process, inclusive-language checks, salary-transparency rules, channel-specific ad formats for LinkedIn, job boards, and short-form blasts, and eight prompts you can run today.

Here is the short version: write a one-paragraph hook, six outcome-based responsibilities each tied to a measurable result, must-have and nice-to-have skills split into separate lists, an honest salary range, and one clear application step. Draft it with AI from a structured brief, run an inclusive-language pass, and get human sign-off before anything goes live.

What a job description is, and how a job ad differs

A job description is an internal document. It defines the role: its purpose, duties, reporting lines, success metrics, working conditions, and pay band. It is the reference you pull out at performance reviews, in promotion conversations, and when a dispute arises about who owns what. A job ad (or job posting) is the external marketing asset you build from that internal document, trimmed and rewritten for the channel where it runs. A job specification is the narrowest of the three: just the skills, qualifications, and experience required.

Conflating the three is how bad ads happen. Paste the internal description onto LinkedIn and you get a wall of compliance language. Write the ad first with no internal anchor, and every interviewer invents their own version of the role. The professional pattern is to write the internal description once, then derive every ad from it.

Titles deserve special care. According to Appcast data cited by the recruiting blog Ongig, job titles of 50 to 60 characters perform 30 to 40 percent better than shorter or longer ones, and Indeed surveys found 36 percent of seekers search by the exact standard job title. "Customer Success Wizard" wins you nothing; "Senior Customer Success Manager (SaaS, Remote EU)" matches how people actually search.

Why the stakes are higher in 2026

The numbers behind a weak job ad have stopped being abstract:

• 44 days. The average US time-to-fill is now 44 days, up 33 percent from 33 days in 2021, according to SHRM's 2025 recruiting benchmarking. Every vague requirement that widens the applicant pool but confuses qualified people adds days.
• $4,700 per hire. SHRM puts average cost-per-hire for a non-executive US role at $4,700. A rewritten ad that fills the role two weeks faster pays for the time invested many times over.
• 6.1 percent apply rate. Appcast's 2025 Recruitment Marketing Benchmark Report (1,300+ US employers, 379 million job-ad clicks, 30 million applications) shows apply rates ended 2024 at 6.1 percent, having climbed 35 percent through the year. Applicants have more choice; clarity is what separates you.
• 47 percent want the salary up front. CareerPlug's 2025 Candidate Experience Report found 47 percent of job seekers expect to learn about salary before applying. Hiding pay now filters out exactly the confident, in-demand people you want.
• 26 to 43 percent. SHRM also reports AI use in HR jumped from 26 percent to 43 percent in a single year. Of HR teams using it, 89 percent say it saves time, but only 17 percent call their implementation highly successful. The winners are not the teams with AI access; they are the teams with a process around it.
• A quality edge. LinkedIn's Future of Recruiting 2025 survey of 1,000+ talent professionals found 37 percent already experimenting with generative AI, and heavy users of AI-assisted messaging were 9 percent more likely to make a quality hire.

If headcount planning sits inside your 2026 business plan, every one of those numbers compounds. Now the workflow.

How to write a job description with AI in 7 steps

Step 1: Run a 20-minute role intake first

AI can only draft what you give it. Interview the hiring manager with five questions: What will this person own in 90 days? What does success look like at six months? Who do they report to? Which three tools will they live in? What salary band is approved? The answers become the model's raw material and your defense against generic output.

ROLE INTAKE BRIEF
Role title (standard, searchable): 
Team and reporting line: 
Location and setup (remote / hybrid / onsite, time zones covered): 
Mission in one sentence (why this role exists now): 
90-day outcomes (3 bullets, each with a number): 
Six-month success metrics: 
Day-to-day reality (top 5 recurring tasks): 
Must-have skills (max 4): 
Nice-to-have skills (max 4): 
Tools used daily: 
Salary band and any transparency rules that apply: 
Application process and assessment steps: 

Step 2: Prompt with the brief, never with a job title

The biggest mistake teams make is typing "write a job description for a marketing manager" and publishing whatever comes back. Feed the full brief instead: the model's job is phrasing and structure, yours is facts.

Act as a talent-acquisition copywriter. Using the brief below, write a job
description for [ROLE] at a [TYPE OF COMPANY, e.g., 12-person Shopify
store / B2B SaaS startup]: a two-sentence hook, six outcome-based
responsibilities, must-have vs nice-to-have skills in separate lists,
benefits, and the salary range exactly as given. Plain language, roughly
8th-grade readability, inclusive wording, no corporate filler. Do not
invent requirements that are not in the brief.

[PASTE INTAKE BRIEF]

Step 3: Make every responsibility an outcome, not a duty

Rewrite any line that starts with "responsible for." The pattern: action verb plus scope plus measurable result. "Manage social media" becomes "Own organic social across Instagram and TikTok; grow engaged reach 15 percent quarter over quarter." Six lines is enough; a duty with no attachable outcome probably does not belong in the ad.

Step 4: Run a separate inclusive-language pass

Do not ask the model to "be inclusive" inside the drafting prompt; run a dedicated second pass so every change is visible.

Review the job description below for inclusive language. Flag: gendered or
aggressive wording (rockstar, ninja, dominant, aggressive), credential
inflation where a skill test would do, jargon that blocks career changers,
 ableist phrases, and any requirement not truly required. Return a table:
original phrase, concern, suggested rewrite. Change nothing silently.

[PASTE DRAFT]

Skills-based phrasing widens reach too: "portfolio with paid-social results" welcomes people whose experience is real even when their titles are not.

Step 5: Decide on salary transparency deliberately

With 47 percent of candidates expecting pay information before applying, the 2026 default is to publish a range. A growing number of US states require it, and the EU pay transparency directive takes effect across member states in 2026. Legitimate exceptions: confidential executive searches and bands still being approved. If you must omit a figure, say when you will share it, because a comically wide range ("$40k–$120k") costs more trust than silence.

Step 6: Format so job boards and applicant tracking systems can read it

Automated parsers and job-board search reward boring structure. Standard job title in the title field, plain text with conventional headings (About the role, Responsibilities, Requirements, Benefits, Salary), no tables or logos in the body, and exactly one application method. Duplicate apply paths confuse parsers and people.

Step 7: Version, check for hallucinations, and get sign-off

Before publishing, verify every credential, license, and legal claim the model produced; AI occasionally invents certification requirements for regulated roles, and the only reliable filter is a human who knows the field. Save the final text as version 1.0 with today's date, log the channels, and calendar a six-month audit. Ads drift out of date quietly; salary bands and tool stacks drift fastest.

This exact chain, structured brief, draft, language pass, KPI extraction, channel variants, and a human checkpoint at the end, is what ArWriter is built for. Plans start at $4.99/month, less than one application of the average $4,700 cost of a mis-hire, and the AI writing toolset covers every document in this workflow.

One-shot AI generator vs. a structured AI workflow

Free one-shot generators have their place: they beat a blank page. But compare what you actually get.

The generator writes words. The workflow produces the document that fills the role.

How Daniel cut time-to-fill from 44 days to 26 at a Manchester logistics startup

Daniel Osei runs hiring at a 35-person logistics startup in Manchester. Early in 2026 he was staring at the industry's average, 44 days to fill a role, while trying to hire six people in one quarter: two operations coordinators, a warehouse shift lead, a finance analyst, and two customer-service representatives.

His old ads were internal descriptions pasted into Indeed: duty lists, no salary, and titles like "Ops Jedi." The rewrite took a different route. He ran a 20-minute intake with each hiring manager, drafted every ad with AI from those briefs, converted each duty line into an outcome with a metric, published real salary bands, and split every posting into a LinkedIn version, a job-board version, and a 50-word blast for the local WhatsApp and Telegram groups where Manchester logistics people actually talk.

The results were quick. The operations-coordinator ad, previously pulling about 41 applications in three weeks with maybe 8 plausible ones, drew 127 applications with 39 plausible ones in ten days. Qualified-applicant volume per ad rose roughly threefold across the six roles. Average time-to-fill for the quarter landed at 26 days against the 44-day benchmark, and because two roles closed before paid job-board spend even kicked in, his blended cost per hire came in about £1,900 under budget. The biggest surprise, he says, was the rejection side: clearer ads meant fewer wrong-fit applicants, and the candidate email templates he set up afterward kept the ones he did not hire from churning in unanswered. He now keeps all six master descriptions in one folder and regenerates channel variants in minutes whenever a role reopens.

One role, three formats: LinkedIn, job board, and short-form blast

Most teams write one ad and paste it into three very different channels, then wonder why LinkedIn feels flat. Each channel has its own grammar:

• LinkedIn post (1,200–1,500 characters): hook first line, since only the first two lines show before "see more"; the mission sentence; three outcome bullets; salary; one application step. Written for the feed, not the listing page. If the hiring manager reposts it, vary the hook, never the facts.
• Job-board long-form (Indeed, ZipRecruiter, national boards): keyword-mapped and structured, standard headings, because board search matches on plain phrases like "customer service representative" and "order management."
• Short-form blast (50–60 words): for WhatsApp communities, Telegram channels, Discord servers, and Slack groups, which is how a large share of job movement actually spreads in Southeast Asia and in trade communities everywhere. Role, city or remote scope, one attraction, salary if appropriate, link, contact.

Treat the three variants like siblings, not clones: same numbers, same promises, different opening lines matched to how each channel is actually read.

Generate all three from the master description so the facts never drift apart:

Convert the job description below into three ad formats:
1. LinkedIn post, 120-150 words, hook in the first line, three outcome
   bullets, salary, one clear application step, no hashtag spam.
2. Job-board listing mapped to standard search phrases for this role,
   with plain section headings: About the role, Responsibilities,
   Requirements, Salary, How to apply.
3. Short-form blast of 50-60 words for messaging apps: role, location or
   remote scope, the single most attractive fact, salary if provided,
   link and contact. Keep every factual claim identical across all
   three versions.

[PASTE MASTER JOB DESCRIPTION]

Agencies hiring for client-facing delivery roles can extend the same discipline to reporting after the hire: standardized job documents pair naturally with automated monthly client reports, since both run on the same brief-draft-review loop.

Should you publish the salary range?

The 2026 answer is yes by default, with narrow exceptions. The case for: 47 percent of candidates expect it before applying (CareerPlug), ranges are already mandatory for many US postings, EU rules arriving in 2026 push the same way, and ads with credible ranges filter out mismatched applicants before anyone spends an interview slot. The mechanics that keep a range honest:

1. Publish the band you can actually pay, not a decoy stretch number.
2. State what drives placement in the range (experience, demonstrated skills, location factor).
3. Keep the same band across all channels; candidates compare.
4. If the role is confidential or the band is mid-approval, name the date you will share it.

Tie every responsibility to a KPI

The fastest upgrade to any draft is attaching a number to each line. Give each responsibility a measurable indicator and a rough weight, and keep the weights summing to 100 percent:

• "Own organic social across two channels — engaged reach +15 percent per quarter — 20 percent."
• "Resolve tier-1 tickets within one business day, CSAT at 4.5 or better — 30 percent."
• "Close two new mid-market accounts per quarter at or above target ACV — 30 percent."
• "Reconcile monthly accounts payable with zero late-payment penalties — 20 percent."

This does two things competitors' templates do not. In the ad, it signals to serious candidates exactly how success is judged, which attracts people who like measurable jobs. After the hire, the same lines become the 90-day review template, so the role starts measured the way it was sold. Review the weights with the hiring manager before posting: if everything weighs 25 percent, nothing does, and a lopsided weighting usually means the role is really two roles.

Extract every responsibility line from the job description below and return
a table with four columns: responsibility, measurable KPI, target number or
direction, weight in percent. Weights must sum to 100. If a line cannot be
measured, rewrite it so it can.

[PASTE DRAFT]

What AI gets wrong, and the checks that catch it

Generative drafting is fast, and its failure modes are consistent. Four to watch:

1. Invented credentials. Models sometimes add certification or license requirements that sound plausible for a regulated role but do not exist or do not apply in your jurisdiction. Verify against your industry regulator's published requirements, or delete.
2. Stale legal claims. Pay-transparency and employment rules change monthly somewhere. Treat any legal statement in a draft as unverified until you check current rules for the posting's location.
3. Cloned boilerplate. Ask ten tools the same vague prompt and you get ten siblings of the same generic ad, including perks you do not offer. Everything in the final ad must trace back to your intake brief.
4. Title inflation. AI happily upgrades "coordinator" to "director" because it reads better. Titles set pay expectations, internal leveling, and search matching; keep them boring and accurate.

The final check is always a named human who signs off on facts, pay, and legal claims before the ad goes live. That signature line is what separates teams reporting real time savings from the 83 percent of SHRM-surveyed AI adopters who stop short of calling their rollout a success.

Ready-to-copy prompts and ad templates

Two prompts appeared inside the steps above (the master drafting prompt and the inclusive-language pass). Here are five more that complete the set. Replace the bracketed fields, keep the facts from your brief, and run each as its own message so you can audit every change.

LinkedIn post skeleton:

[Hook: the one problem this role owns, in one line]

We're [company, one clause on what you do] hiring a [standard job title]
to [mission sentence from the brief].

You'll:
- [Outcome 1 with metric]
- [Outcome 2 with metric]
- [Outcome 3 with metric]

Pay: [range] [placement driver, one clause].
Location: [remote scope / city / time zones].

Apply: [single method + what happens next + timeline].

Short-form blast for messaging apps:

Hiring: [standard job title]
[City / remote scope]
[Math of the mission in plain words]
Pay: [range if public]
Full ad: [link]
Questions: [contact]

Job-board listing section order:

Title: [standard searchable job title] + [one modifier: team or scope]
About the role: 2-3 sentences, mission first
Responsibilities: 6 outcome lines with metrics
Requirements: max 4 must-haves
Nice to have: max 4
Salary: [band + placement driver]
Benefits: [only perks you actually offer]
How to apply: [one method + process + timeline]

Rewrite prompt for an existing weak ad:

Here is our current job ad. Rewrite it against the brief below: kill duty
language, convert each responsibility to an outcome with a metric, split
must-have from nice-to-have, add the salary range verbatim, and cut total
length by one third. Keep every fact that is still true; flag facts you
cannot verify.

[CURRENT AD]
[BRIEF]

Interviewer-alignment prompt, the step almost everyone skips:

Based on this final job description, produce a one-page interviewer
guide: five competency areas drawn from the responsibilities, one
suggested question per area, and the KPI each area maps to, so every
interviewer tests the same role.

Job description launch checklist

Before any ad goes live, run this list once:

1. Intake brief completed and approved by the hiring manager.
2. Standard, searchable job title in the title field.
3. Six outcome-based responsibilities, each with a metric.
4. Inclusive-language pass run and its changes reviewed.
5. Salary decision made deliberately and applied consistently.
6. Every credential and legal claim verified by a human.
7. Three channel variants generated from one master document.
8. One application method, with the process and timeline stated.
9. Version 1.0 saved, dated, and its channels logged.
10. Six-month audit scheduled, and once the role is filled, prepare the employee handbook before hire number one walks in.

Frequently Asked Questions

What is a job description?

A job description is an internal document that defines a role's purpose, duties, reporting lines, success metrics, and pay band. It anchors performance reviews and hiring decisions. A job ad is the shorter, external version derived from it, written to attract candidates on a specific channel. Writing the internal description first keeps every ad consistent.

What should you include in a job description?

Include a mission sentence, six outcome-based responsibilities with measurable results, a maximum of four must-have skills, nice-to-have skills listed separately, the salary range, real benefits, location and remote scope, reporting line, and one clear application step. Keep it under roughly 700 words, and format with plain headings so job-board systems can parse it.

How do you write a job description for a new position?

Start from outcomes, not history, because no incumbent exists to copy from. Ask the hiring manager what the business needs in 90 days and at six months, convert each answer into a measurable responsibility, and decide the reporting line and pay band before drafting. Draft with AI from that brief, then validate with everyone who will interact with the role.

How do you write a job description on LinkedIn?

Put the hook in the first two lines, because the feed truncates the rest. Follow with the mission, three outcome bullets with metrics, the salary range, remote scope, and one application step. Keep the full post under about 1,500 characters. Use the standard job title in any linked listing so LinkedIn's search matches it, and skip hashtag piles.

Why do job descriptions matter?

They set expectations before day one, which is why clear ads filter mismatched applicants before anyone spends interview time. SHRM's 2025 benchmarking puts average cost-per-hire at $4,700 and time-to-fill at 44 days, so every avoided mis-hire and every shortened week is real money. After the hire, the same document becomes the baseline for reviews and promotion cases.

Are AI job description generators free to use?

Many are free for a first draft. The catch is what happens after: one-shot output usually needs an hour of repair across responsibilities, inclusive language, salary handling, and channel formatting. Teams that draft regularly tend to move to a structured tool such as ArWriter, from $4.99/month, that manages the whole chain.

Can AI write a job description that job boards and ATS software parse correctly?

Yes, if you enforce plain formatting: standard searchable job title, no internal jargon, plain-text sections with conventional headings, no tables or embedded graphics in duty lines, and one application method. AI output pasted with decorative formatting is what breaks parsing. The structured workflow in this guide includes a formatting step for exactly this reason.

Should you include salary in a job posting?

In 2026, yes by default. CareerPlug found 47 percent of candidates expect salary information before applying, several US states require ranges, and EU rules push the same way. Publish a band you will genuinely pay, explain what drives placement, and keep it consistent across channels. Omit it only for confidential searches, then say when you will share it.

Sources

• SHRM Talent Acquisition news and research
• LinkedIn Future of Recruiting 2025
• Appcast 2025 Recruitment Marketing Benchmark Report
• CareerPlug Candidate Experience Statistics
• Gallup on preventable employee turnover

Try ArWriter Today

The teams winning at AI-assisted hiring are not the ones with the fanciest prompts. They run a short, repeatable chain: a real brief, a drafted description, a language pass, KPI-mapped responsibilities, channel variants, and a human signature. ArWriter runs that chain from $4.99/month, with the prompt library and writing tools to keep every hiring document consistent, from the first ad to the offer email.

Start free →

=== Article 2 ===