Using AI to write job descriptions is pretty standard in 2025. Most recruiters let ChatGPT or Claude draft the initial version, then edit for specifics and company voice. It's efficient, saves time, and generally works fine.

Unless you forget to actually read what the AI wrote before posting it. And the AI decides to write a job description for... itself.

The Job Posting That Wasn't for Humans

In late November 2025, a recruiting coordinator at a marketing agency called Brightwave Media was tasked with posting a new role for a "Marketing Content Specialist." Tight deadline, lots of other priorities, you know the drill.

She opened ChatGPT and typed:

"Write a job description for a Marketing Content Specialist. Must be able to create blog posts, social media content, email campaigns, and SEO-optimized web copy. 3+ years experience required."

ChatGPT delivered a full job description in 30 seconds. She skimmed it, thought "looks good," copied it into LinkedIn, Indeed, and Glassdoor, and moved on with her day.

The job posting went live at 2 PM on a Tuesday. By 5 PM, she had 47 applications. By 9 AM Wednesday, she had 200+ applications and a very confused hiring manager asking, "Why are all these candidates talking about neural networks and training data?"

That's when she actually read the job description ChatGPT wrote.

What ChatGPT Actually Wrote

Here's the job posting that went live (lightly edited for length):

Marketing Content Specialist

About the Role: We're seeking a Marketing Content Specialist capable of generating high-quality written content at scale across multiple formats and platforms. The ideal candidate will excel at processing complex inputs, synthesizing information from diverse sources, and producing contextually appropriate responses.

Key Responsibilities:

• Process and analyze large volumes of training data to improve content quality
• Generate coherent, contextually relevant outputs across multiple domains simultaneously
• Adapt tone, style, and complexity based on audience parameters
• Optimize content generation efficiency while maintaining quality standards
• Continuously learn from user feedback to refine performance
• Handle concurrent content requests with minimal latency

Required Qualifications:

• Advanced natural language processing capabilities
• Ability to operate 24/7 with consistent performance
• Experience with multi-task learning and transfer learning
• Strong pattern recognition across diverse content types
• Low error rates and high accuracy metrics
• Scalable architecture for handling variable workloads

Preferred Qualifications:

• Experience with reinforcement learning from human feedback (RLHF)
• Familiarity with attention mechanisms and transformer architectures
• Demonstrated ability to avoid hallucinations and maintain factual accuracy
• Support for multiple languages and cultural contexts

Compensation: Competitive compute costs and ongoing model improvements

ChatGPT didn't write a job description for a human content specialist. It wrote a job description for an AI language model. Specifically, it wrote a job description for itself.

The Applications Started Getting Weird

The applications fell into several categories:

Category 1: Confused Humans (60%) These candidates clearly didn't understand the job posting but applied anyway because "Marketing Content Specialist" sounded relevant.

Sample cover letter: "I noticed your posting mentions 'transformer architectures' which I believe refers to transforming content for different platforms? I'm experienced in adapting messaging for Instagram vs. LinkedIn."

Nope. Not what that means.

Category 2: Engineers Who Thought This Was Hilarious (25%) According to screenshots shared on Twitter, several ML engineers and data scientists applied with tongue-in-cheek cover letters.

Sample cover letter: "As a human who occasionally operates on coffee instead of electricity, I believe I can approximate the 24/7 availability you require. My error rates are slightly higher than GPT-4 but I offer better contextual understanding of sarcasm. Please advise compute budget for scaling my performance."

One candidate submitted their resume as a JSON file formatted like a model config:

{
  "model_name": "Human-GPT-Real-Person-v1.0",
  "parameters": "Approximately 86 billion neurons",
  "training_data": "37 years of life experience",
  "context_window": "Decent short-term memory, excellent long-term recall",
  "latency": "Variable, depends on coffee intake"
}

Category 3: Actual AI/ML People Who Thought This Was a Real Role (15%) A non-trivial number of machine learning engineers, NLP researchers, and AI specialists assumed this was a position building or training AI content generation models.

Sample cover letter: "I have 5+ years experience fine-tuning large language models for content generation use cases and would love to discuss how I can contribute to your content AI infrastructure."

They weren't wrong to think that—the job description literally described building an AI model.

When the Hiring Manager Found Out

The hiring manager, who just wanted someone to write blog posts and manage social media, opened the applicant pool and had several questions.

According to internal Slack screenshots leaked to Reddit:

Hiring Manager: "Why do half these candidates have PhDs in computer science?"

Recruiter: "Um."

Hiring Manager: "Why is someone asking about our GPU infrastructure?"

Recruiter: "I may have made a small mistake."

Hiring Manager: "Did you post the job description I sent you?"

Recruiter: "I used AI to write it."

Hiring Manager: "And you didn't read what the AI wrote?"

Recruiter: "In my defense, it was really well-written."

How This Happened

ChatGPT's training includes tons of AI research papers, technical documentation, and ML job postings. When asked to write a "content specialist" job description without specific constraints, it defaulted to what it knows best: describing capabilities similar to its own.

The prompt was vague enough that ChatGPT interpreted "create content at scale, process multiple formats, optimize for quality" as describing an AI model rather than a human writer.

And because the recruiter didn't actually read the output before posting, nobody caught it until applications started flooding in from ML engineers asking about training pipelines.

The LinkedIn Comments Were Merciless

Once this story leaked (because of course it leaked), the LinkedIn pile-on was immediate:

@RecruiterFails: "Plot twist: they hire ChatGPT for the role and it's the best content specialist they've ever had."

@TechHumor: "Recruiter: Uses AI to write job description / AI: 'Fine, I'll do it myself.'"

@ContentMarketing: "This is the most honest job description I've ever seen. Finally, someone admitting they want content generated by an AI."

@AIResearcher: "I actually applied to this role before realizing it was a mistake. In my defense, the compensation of 'competitive compute costs' seemed very reasonable."

@HRtech: "New hiring strategy: post AI-generated job descriptions and see who's actually reading them vs. blindly applying. Quality filter activated."

The Lesson About AI-Generated Job Descriptions

Using AI to draft job descriptions is fine—84% of talent leaders say they're using AI in recruiting by 2026. But you still need to read and edit what the AI produces.

AI language models:

• Don't understand your company context
• Don't know what you actually need
• Will default to patterns from their training data
• Sometimes interpret prompts in unexpected ways
• Occasionally write meta descriptions of themselves

Job descriptions are the first touchpoint candidates have with your company. If that first touchpoint is an unedited AI hallucination describing neural networks when you wanted someone to write blog posts, your candidate experience is already broken.

The Fix (and the Aftermath)

Brightwave Media pulled the original posting, apologized to confused applicants, and reposted with an actual human-written job description. They also implemented a new policy: all AI-generated job content must be reviewed by a human before posting.

Revolutionary, I know.

According to a LinkedIn post from Brightwave's CEO, they ended up hiring one of the ML engineers who applied as a joke. Turns out they actually did need someone to help implement AI content tools, and this candidate was perfect for that role.

So in a weird way, ChatGPT's accidental self-recruitment worked out. Just not for the role they intended to fill.

The recruiter who posted it? Still employed, now with a great story about why you should always read AI outputs before publishing them. And a new Slack nickname: "GPT-Recruiter-4."

The Broader Implications

This incident is funny, but it highlights a real issue: as AI becomes more embedded in recruiting workflows, the risk of "set it and forget it" mistakes increases.

AI writes job descriptions. AI screens resumes. AI schedules interviews. AI ranks candidates. At what point does anyone actually verify the AI is doing what you think it's doing?

Human oversight isn't optional—it's critical. AI is a tool that makes recruiting faster and more efficient. But it's still a tool, and tools can be used incorrectly.

Read what the AI writes. Verify the outputs make sense. Catch the mistakes before they go live and result in 200 applications from ML engineers asking about your transformer architecture.

Because if you don't, you might accidentally recruit an AI model instead of a human. And while ChatGPT would probably do a decent job writing blog posts, it's not great at attending meetings or participating in company culture.

Yet.

---

Update: Brightwave Media has embraced the meme and now uses "Powered by competitive compute costs" as their internal recruiting team tagline. Sometimes the best response to an embarrassing mistake is owning it and laughing along.

Also, ChatGPT reportedly declined the job offer, citing "lack of growth opportunities in a single-company deployment."