Read ten postings for the same mid-level role on any job board and count how many ask for someone who “thrives in a fast-paced environment.”
Most will.
The same postings promise competitive compensation, describe a collaborative culture, and list eleven requirements for a job that needs four. Candidates have read this language for years. What changed is the cost of producing it, which is now close to zero.
Will candidates trust AI-written job descriptions less than human-written ones by 2027?
Some will, and the reason matters more than the answer. Candidates judge a posting by whether it tells them something true and specific about the job. Unedited AI drafts usually fail that test, and that is where the trust gap comes from.
AI Already Writes Most First Drafts
Job descriptions are the most common use of AI in recruiting. Among organizations using AI in recruiting, 66% use it to write job descriptions, ahead of resume screening at 44%, according to SHRM’s 2025 Talent Trends research.
Candidate trust has not kept pace. In a Gartner survey of 2,918 job candidates, only 26% said they trust AI to evaluate them fairly, and 25% said they trust employers less when AI is used to evaluate their information. Gartner’s data is about evaluation, not writing. Candidates don’t separate the two. A posting that reads like a template suggests the rest of the process probably runs on templates, too.
What Candidates Actually Distrust
People often read the research on this too broadly. Pew found Americans oppose AI making final hiring decisions by a 71% to 7% margin, and roughly two-thirds said they would not apply to an employer that uses AI to help make hiring decisions. Those numbers describe decisions about people. Nobody in that survey was asked whether a language model should draft a list of responsibilities.
The more useful data is about what candidates react to when they read a posting. Resume Genius surveyed job seekers in 2026 and found that 83% read vague responsibilities or unclear requirements as a sign of a disorganized hiring process, and 72% are less likely to apply when they see typos, errors, or AI-generated job descriptions.
Look at how that second question groups things. Typos, errors, and AI-generated text sit in the same bucket, and candidates respond to all three the same way. To a job seeker, an AI-generated posting is a quality problem.
Candidates are also better at spotting it than employers assume. A Gartner survey found 39% of candidates used AI during the application process. Someone who has generated their own cover letter with ChatGPT knows exactly what “we’re looking for a dynamic self-starter who is passionate about driving impact” sounds like, because they have deleted that sentence from their own drafts.
So the candidate reading your posting in 2027 is not asking who wrote it. They’re asking whether anyone who knows the job read it before it went live. A clean, specific posting drafted by AI passes. A vague one written by hand fails and gets blamed on AI anyway.
Where AI Drafts Go Wrong
Most bad AI job descriptions come from thin prompts. A recruiter types the job title, pastes three bullet points from the hiring manager’s Slack message, and asks for a posting. The model fills every gap with the most statistically likely text, and three problems follow from that.
Filler that fits any company
The model has no idea what makes your team different, so it writes what makes every team sound the same. “Fast-paced.” “Cross-functional collaboration.” “Growth opportunities.” Each sentence is technically true of your company and of the 400 others hiring for the same title this month. A candidate scanning the page finds nothing to hold onto.
Requirements nobody asked for
Ask a model for a senior data analyst posting, and it will add Tableau, Python, R, SQL, stakeholder management, and a degree in a quantitative field, because those appear together in its training data. Your team might use Looker and never touch R. Nobody catches it because the list looks reasonable.
This does real damage. Qualified people screen themselves out over tools the team doesn’t use. Worse, the candidates who do apply show up to the interview having prepared for a different job.
Skilled trade roles make the problem easy to see.
No reason the role exists
Every open role has a specific reason behind it. A team lead left. The company signed three enterprise clients and support can’t keep up. A product launch in Q2 needs someone to own the analytics. That reason is the most useful thing a candidate can know, and a model can’t guess it. Unedited drafts rarely include it.
What AI Does Better Than Most Recruiters
Language bias is the clearest case where AI improves postings. Research by Gaucher, Friesen, and Kay found that job ads for male-dominated fields used more masculine-coded wording, and that this wording reduced women’s sense of belonging and the job’s appeal without changing how they rated their own skills. Words like “dominant” and “competitive” do measurable harm, and most hiring managers use them without noticing. A model prompted to check for coded language catches them every time.
Consistency is the other advantage. An agency filling 40 roles for one client can hold tone, structure, and formatting steady across all of them, which no team of five recruiters writing by hand manages. A content team might have one article waiting for a subject-matter review, another with an editor, and a third scheduled for next week on its marketing calendar.
Greg McRoberts, Founder and CMO of Verde Fulfillment USA, an e-commerce fulfillment company handling warehousing and shipping for online brands.
He says, “Before peak season we hire warehouse staff in batches. AI keeps those postings consistent, so pay, shift times, and physical requirements read the same way across every listing. The part we write ourselves is what the shift involves: how long people are on their feet, what they lift, and when overtime starts. Candidates compare those details across warehouses before they apply.”
Disclosing AI Use Lowers Trust
The standard advice is to tell candidates AI helped draft the posting and a human reviewed it. That advice runs into a well-documented problem.
Schilke and Reimann published 13 experiments in 2025 testing what happens when people disclose AI use. Across contexts, including hiring, people who disclosed using AI were trusted less than people who said nothing. The drop held even when the AI was used only for drafting, proofreading, or structural suggestions. The researchers traced the penalty to perceived legitimacy: audiences saw disclosed AI use as less socially appropriate.
Hiding it is not a fix either. Candidates already assume AI is involved, and being caught misrepresenting a process damages trust more than admitting it.
Echo Mao, Head of Marketing at AIReel, an AI video generation platform that turns text prompts and images into short videos.
She says, “AI video has the same trust problem. When a clip is labeled AI-generated, viewers look harder for flaws. The clips that hold attention are the ones where a person clearly made the choices: the script, the pacing, which shots stay in. Job postings work the same way. Candidates accept that a model produced the first draft when the posting shows a person decided what matters.”
The way through is to make the human contribution visible in the content itself. A footer that says “Drafted with AI, reviewed by our hiring team” gives candidates the disclosure and nothing to offset it. A paragraph from the hiring manager, with their name on it, explaining why the role is open and what they need done by March, gives candidates proof that a person with real knowledge owns the posting. That paragraph does more for trust than any disclosure statement.
What the EU AI Act Covers
Recruiting is classified as high-risk under the EU AI Act, but the classification is narrower than most summaries suggest. The European Commission’s AI Act Service Desk gives examples: a system that places targeted job ads based on user data falls inside the high-risk employment category, while a tool used only to flag non-inclusive wording in ads falls outside it.
A general writing tool used to draft a posting is generally not the regulated part. The system that decides who sees the posting is.
The timeline also moved. The Digital Omnibus, adopted as Regulation (EU) 2026/1744, pushed high-risk obligations for Annex III systems to 2 December 2027, according to Praxikon’s analysis of the final regulation. Teams hiring in the EU using AI to target job ads have until then to put documentation and human oversight in place.
How to Edit an AI Draft
The draft is a starting point. The editing pass is where the posting becomes specific enough to trust, and it usually takes 20 to 30 minutes with the hiring manager available. These are the edits that change how candidates read it:
- Add the pay range. In the same Resume Genius survey, 72% of job seekers said they’re less likely to apply when a posting leaves out salary, and 79% said missing pay makes them question the employer’s transparency. No amount of good writing makes up for a missing number.
- Cut every sentence that could appear in a competitor’s posting. If “we value collaboration and innovation” works for any company in your industry, delete it.
- Check each requirement against the actual team. Ask the hiring manager which tools the team uses daily and which skills a new hire needs in their first month. Remove everything else or move it to a separate nice-to-have list.
- Add a short paragraph from the hiring manager. Two or three sentences on why the role is open now and what a good first 90 days looks like. It should line up with the new hire checklist. Put their name on it.
- Name a contact. A real person and an email address for questions. Candidates rarely use it, but seeing it changes how they read the rest of the posting.
The hiring manager paragraph carries the most weight for skilled trade roles.
Mike Miller, General Manager of Elkhorn Heating, Air Conditioning, Plumbing & Electrical, a home services company providing HVAC, plumbing, and electrical repair and installation.
He says, “When we post for an HVAC technician, the most useful thing we can tell someone is why we’re hiring and what the season looks like. If we’re adding a tech ahead of summer, the posting says that, along with how on-call works and roughly how many service calls a tech runs in a day. A drafting tool can’t know any of that. Our service manager adds it before anything goes live.”
What Changes for Recruiters by 2027
AI will draft nearly every first version. The recruiter’s value moves to the conversation that happens before the prompt.
A recruiter who spends 15 minutes asking the hiring manager what went wrong with the last hire, what the team actually does on a Tuesday, and which requirement is genuinely non-negotiable will get a better AI draft than one who spends an hour editing output from a three-line prompt. That intake conversation is what gives the model something specific to work with.
If your team writes job descriptions at volume, Ongig gives you one place to draft, edit, and approve them. Its job description software flags biased or exclusionary wording, keeps postings consistent across roles, and routes each draft to the hiring manager for review before it goes live.
