Looking for the job post audit or my free LLM prompt guide? Scroll to the bottom of the page!
On the list of things recruiters want to automate most, job post writing is usually in the top 5 depending on how much the vendor wants to sell you some AI job post writing bot. After training people to write job posts for the last 8 years, I am not surprised this is so popular. Most recruiters don’t think they’re good writers. Of course, they think AI that makes things sound good is better at writing than they are.
As a writer, I actually think most recruiters are good at writing about work - there’s just a block between their brain and their hands. Too many recruiters were trained on overly formal communication and somewhere that translated to, “you can’t just write down what you say.” But you can and should. If you know what to say, you know what to write for a good job post.
A good job post is written like someone is explaining things to you. It’s human. It’s straightforward. It gives the insider information that only someone in the job can tell you. Ultimately, it helps you answer 2 questions: Do I want to do this? Can I do this? Most of all, it’s not generic like this junk LinkedIn AI job post. And I say AI lightly because “intelligent” is a leap…
My AI Job Post Test: The Methodology
Anyone can click a button, generate a generic job post, and sift through a pile of unqualified people for a week. AI creates room for people to differentiate with their voices. That’s why we shouldn’t be using AI to write job postings in this one-dimensional plug and play system anymore. It’s not “here’s the title, write the job post” because AI really sucks at that. That’s the first thing I learned running my own AI job post test.
My methodology was simple: find 5 different AI tools to write 5 different job posts. Evaluate these AI samples vs 5 human samples written during my last job post writing training. I scored each job post to see which group has the higher score in each category - AI or Human. To see where they struggle or succeed.
The categories of my audit were
- Job Title: When you search this title, does it return job posts for roles that require similar skills?
- Template: Does it follow template and brand guidelines?
- Tone: Does it sound human? Does it use buzzwords?
- Bullets: How many?
There’s more, but you get the gist.
Who Is Better At Writing Job Posts - AI Or Humans?
I learned a lot by seeing them in contrast with student samples from my own training, where I taught them what good means and how to generate that content themselves.
- What AI is good at:
- Title analysis. I’m biased. I built the tool that does this - job title generator - and now I teach teams to use it. But it helps get a better idea of how skills align to titles in a way that you would never think about if you’ve worked at a company for a long time and it has “always been that way.”
- Template. Replicability is an AI strong suit. It’s great at providing well formatted content if you give the tool a template.
- Clarification. How do you go from 800 words to 300? The prompts below will help with these simple checks for job clarity.
- Grammar. Spell check is your friend, shit managers… I mean shift managers.
- What humans are good at:
- Tone. In fairness, AI did well with a lot of direction and some learning samples. But the thing the people did even better was insert all the details like scope and scale. The words they chose were more contextual with the job and the team so it would feel familiar to someone in the industry.
- Bullets. AI loves to bullet bomb a job post. It naturally went toward all bullets unless explicitly prompted not to. My students know how many bullets go into a job post.
- Data Collection. Prompting someone to answer questions on a screen is not as effective as a conversation to get clarity about what a role entails and what requirements matter. Humans lead better strategy sessions than bots.
Knowing this, I recommend a process that teaches recruiters how to write a job post first, then integrates AI prompting (get my favorites below!) and tools into stages for optimization where it can be most impactful. For example, doing a job title analysis before intake with AI then using simple meeting summary AI after the intake meeting to generate the first draft of bullets from the conversation. No HR tech required.
Oh, and never let LinkedIn write your company description. If you ever see this company description online, know this is a cry for help. Save me.
I would NEVER.
Want some help building your AI job post writing process so it generates quality, not crap? Book a meeting.
