Why your AI-written job ad isn't helping

Suppose I have a job to fill and need several people for it, and keeping staff has been the difficult part. I decide to use AI, and my first move is to have it write my Indeed or ZipRecruiter ad.

That feels like a smart use of AI. It isn't.

Everybody has an ad

On its own, an ad is no advantage, because every business has one.

The obvious objection is that the AI version looks better, and it does. It probably reads better than anything I've written myself. The catch is that a thousand other owners are doing the same thing, and their ads look better too.

Economists Emma Wiles and John Horton have studied this in a large field experiment on an online hiring site. Some employers were offered an AI-written first draft of their job post. Those employers were 19% more likely to post a job, and they spent 44% less time writing it.

The number of hires did not go up. The AI posts were more generic and said less about the actual job, so job seekers spent their time on posts that led nowhere.

The researchers' explanation will make sense to anyone who has hired. A long, detailed post used to tell a job seeker the owner was serious, because it took effort to write. Today, anyone can produce a long, polished post in seconds. Length no longer signals anything, and the reader can't tell a serious owner from a lazy one.

AI-written posts spread quickly, given that ChatGPT came out only in late 2022. By October 2023, a study in the journal Patterns estimated that roughly 5% to 10% of LinkedIn job posts from smaller employers already showed AI-assisted writing.

A better ad brings in more of the same candidates. It won't tell you who's going to stay. Using AI to make a more polished version of what everyone else makes doesn't put you ahead. That distinction is what Founder AI is about.

"A better ad brings in more of the same candidates. It won't tell you who's going to stay."

Why do new hires leave in the first month?

You already know how to attract candidates, so finding them isn't the problem. The real problem is that too many turn out to be wrong for the job, or they take it and don't stay.

Most likely, you pitched the job incorrectly.

People quit in every kind of business. The Bureau of Labor Statistics counts how many workers quit each month. In August 2026, 1.9 of every 100 workers quit across all industries. In professional and business services, which include accounting, law and consulting offices, the figure was 2.1. In retail, it was 3.0.

Those are monthly figures. In a five-person office, one quit is 20% of your staff.

Much of that turnover happens early. In a 2018 survey by the recruiting software company Jobvite, 30% of job seekers said they had left a new job within the first 90 days. The top reason, given by 43% of them, was that the day-to-day role wasn't what they expected.

In other words, the ad described one job, and the workday turned out to be another.

What does one early quit cost?

Most owners count only the ad, but the full bill is larger.

The Center for American Progress reviewed 11 research papers on the cost of turnover. For jobs paying under $30,000 a year, the typical cost of replacing someone came to 16% of that person's yearly pay.

That figure covers the ad, the screening, the interviews and the training. It also counts costs that never appear on a receipt. Those include work left undone while the position is empty, mistakes while the new person learns and a drop in everyone else's morale.

Here is a hypothetical example to show the math. Substitute your own numbers as you go.

Suppose you run a 12-person insurance office and hire a front-desk coordinator at $17 an hour for 30 hours a week. In week three, the coordinator doesn't come back.

• $150 to boost the ad on a job board.
• Six hours of your time reading applications and interviewing. At $40 an hour for your time, that comes to $240.
• Three training days, or 24 paid hours at $17. That comes to $408.
• Overtime for the rest of the office to cover the front desk, 20 hours at $25.50 an hour. That comes to $510.

Those costs add up to $1,308 you can actually see. Next, apply the research figure. At 30 hours a week, the coordinator would earn about $26,520 a year, and 16% of that is about $4,240.

On top of that, you have to start over and pay most of those costs again for the next hire.

What does "pitching the job right" look like?

Most job ads promise the top of the scale. If you're the perfect candidate, give your life to the job, work hard and do exactly what you're told, here's what you'll make. It's rarely that clear-cut. Development usually isn't part of the pitch, either. If all you're offering is money, good people will find better pay somewhere else.

Researchers have a name for the honest version: a realistic job preview.

The idea has been studied for decades. A 1998 review of 40 studies in the Academy of Management Journal tied realistic previews to lower turnover, including people quitting on their own. The effect is small, and it isn't a cure-all. A 2011 review suggests why it works at all. Candidates come away believing the company is honest with them.

Below is a general checklist for an honest pitch. It's common practice, not a script, and you'll know better than anyone what belongs on it for your business.

• The pay and how it works in practice: base rate, overtime or commission and when raises happen. Some states and cities, Washington among them, require a pay range in the job ad, so check your state labor department's rules.
• The schedule as it runs week to week: early starts, late finishes, any weekend or on-call time and how far ahead the schedule is posted.
• The hardest part of the job in plain words. That might be a phone that rings all day at the front desk or month-end deadlines for a bookkeeper. For a field-service dispatcher, it's rebuilding the schedule when the first job runs long.
• What the first two weeks look like, including who trains them and for how long.
• What doing the job well looks like at 90 days.
• A way to see the job firsthand, such as a 15-minute walk-through on a real workday before they say yes.

None of that is hard to write. An AI-written ad still won't include it, because the AI knows none of it unless you tell it.

What to build instead of a better ad

Rather than polishing the ad, build a job-fit questionnaire. Every candidate gets the same questions, tied to the real duties of the job. Behind the scenes, Claude helps you write questions about real moments from the workday, so the answers show who's reliable and who understood the job they're signing up for.

For an office assistant, one question might be what they do when two lines ring while a client waits at the desk. Keep every question about the work, not someone's personal life, past pay or how long they plan to stay.

Next, have Claude build a simple page that lays the answers side by side, next to a scoring guide you wrote. The page shows your scores. More important, it shows you the one thing that matters most to you. For me, it's availability and reliability. Right now, everyone needs more money, so you need to see who is truly loyal. Otherwise, they'll work three days for you and start looking for a job that pays more. If one appears, you'll lose them.

Claude never scores or ranks anyone, so the scores on that page are yours.

You read the answers, and you make the call. Candidates are told that AI helped build the questions.

The honest pitch and the questionnaire work as a pair. The pitch tells candidates what the job truly involves, and their answers show you who heard it.

Common questions

Should I use AI to write my job ads?

You can, and the ad will look better, but so will everyone else's. In the Wiles and Horton experiment, AI drafts led to more posted jobs but no more hires, because the posts were generic. If you use AI for the ad, spend your time on the truth about the job, not the wording. The checklist above is a good place to start.

Why do new hires quit in the first month?

Most often, the job wasn't what they were told. That was the top reason in the Jobvite survey, at 43%. Across all industries, 1.9 of every 100 U.S. workers quit in August 2026 alone, according to the Bureau of Labor Statistics. Pitch the job honestly, and then give every candidate the same job-fit questionnaire to judge fit before day one.

Should I put AI skills in a job ad?

For an office job, yes. Hire people who are a B or better with AI, and aim to turn them into an A. Once they reach an A, you have a real employee.

Listing the skill in the ad isn't enough, so test for it. Have Claude help you build a short test so applicants show you their AI skills instead of claiming them. The post "Build an AI skills test for job applicants" explains how.

You won't be the only employer asking, and Indeed's Hiring Lab keeps count. As of September 30, 2026, about 7.3% of U.S. job posts on Indeed mention AI, up from about 3.7% a year earlier.

Not every job is an office job. Someone has to dig the ditches, take out the trash and cook the food. Those are trades, and an AI line doesn't belong in the ad for a cook or a dishwasher. Save it for the office.

Take more than two seconds

When you build this, spend more than two seconds on it. Build something you'll keep, improve and use for every hire.

Major corporations have paid tens of thousands of dollars for this kind of tool over the years. Some well-known assessment platforms start at $10,000 a year and rise with headcount. You can build a version that fits your business for the price of an afternoon.

This project is a prime example of how a founder or business owner can use AI to get ahead.