Text us your comments or topic ideas for future shows.

One in four US employers are now using AI in their hiring. And one in three job posts are not real jobs at all.

On this episode of The KeyHire Small Business Podcast, Corey Harlock sits down with keynote speaker, LinkedIn Top Voice, and bestselling author Kat Kibben to unpack why AI recruiting is really a math problem that most companies are getting wrong.

Kat has spent more than 15 years helping organizations like Zoom, Monster, and Olly write better job posts and navigate hiring with confidence, with work featured in the New York Times, NPR, and Forbes. Kat's core message is simple but powerful: AI cannot find what you cannot define. And most businesses, especially small businesses, have a definition problem. When you never clearly outline what you are looking for, every step that follows, from interview questions to scorecards to the first ninety days, is built on a shaky foundation.

Early in the conversation, Corey and Kat find instant common ground on the number one mistake business owners make when they hire. They do not define the role. Instead, they build what Corey calls a wish list, a job description packed with responsibilities no single person could ever handle. Kat explains that a strong job post is not about industry or seniority. It comes down to four categories of information: aptitude and the tools you use, teamwork and who you work with, who you present ideas to, and autonomy or whether you can do the work on your own. Both admit their own hiring record proves the point, with Kat noting that even as a recruiting expert running a recruiting business, only one in four early hires was the right one.

From there, the discussion moves into how AI is reshaping recruiting and hiring, for better and for worse. Kat ran a head to head test of the major large language models, feeding each one identical information and asking for a job post. The results hallucinated a location that did not exist, invented an inaccurate pay range, and exaggerated requirements back into old school formats, because these tools are trained on decades of flawed job postings. Kat warns that AI is a tone thief, and explains why prioritizing speed over results leads to biased screening, rejection loops, and lawsuits.

The episode also takes a fascinating and slightly alarming turn into candidate harvesting, fake job posts, and the rise of fraudulent candidates. Kat walks through activity now being investigated by the FBI and CIA, where foreign agencies mirror the job postings of well known American companies to collect resumes, then send AI backed candidates into interviews at large organizations to steal entire code bases and client lists. Corey shares his own firsthand experience interviewing three separate candidates who all listed the same three companies, every one of which turned out to be a fake website.

What makes this conversation so valuable is how practical it stays. Kat and Corey agree that great hiring is a discipline, not a gut decision, and that most owners get it right only one in three or one in four times. They dig into how to build interview scorecards from a well defined job post, why you have to define what a good answer sounds like before you ever ask the question, why communication and timeline transparency keep top candidates engaged, and how findable job titles help the right people discover your role.

If you are a small business owner, entrepreneur, or leader trying to use AI in your hiring process without falling into expensive traps, this episode is a must listen. Tune in for a smart, data backed conversation on job postings, AI recruiting, candidate screening, and why defining what you want is still the most important hiring decision you will ever make. Stop grinding, start growing.

KEY TAKEAWAYS

AI cannot find what you cannot define: the job post sits at the bottom of everything, and if you get it wrong, your interview questions, your screening decisions, and your six month performance benchmarks are all built on nothing.

You are writing a wish list, not a job post: a job post exists so that anyone reading it can say a clear yes or a clear no, and if unqualified people keep applying, the post was never clear enough to let them decide.

Four categories make a job post work: the tools you use and how you use them, who you work with, who you present ideas to, and whether you can do the work autonomously. Not years of experience, not degrees, not detail oriented.

AI is a tone thief trained on bad data: in a head to head test of the major models with identical inputs, one invented a location that did not exist, one added an inaccurate pay range, and all of them reformatted requirements back into the old templates they were trained on.

Speed is not a benchmark of success: time to fill measures things you do not control, and when you anchor people to time, they start rejecting anyone scored below an eight. That checkbox looks like efficiency and can get you sued.

Your job post is an attack surface: one in three postings are not real jobs, foreign agencies mirror well known company postings to harvest resumes, and fraudulent candidates have used deepfake video to get hired and walk out with code bases and client lists.

Define what a good answer sounds like before you ask the question: if you skip that step, you will default to whoever communicates most comfortably, which is bias, not assessment.

Findability is free and almost nobody does it: Google your job title alongside the word resume. If the resumes you would hire do not come up, nobody is going to find your role.

LINKS & RESOURCES

Kat Kibben's company, Three Ears Media: https://threeearsmedia.com/

Connect with Kat Kibben on LinkedIn: https://www.linkedin.com/in/katrinakibben/

Connect with Corey Harlock on LinkedIn: https://www.linkedin.com/in/coreyharlock/

Learn more about KeyHire Solutions: https://www.keyhire.solutions

Subscribe on Apple: https://podcasts.apple.com/us/podcast/the-keyhire-small-business-podcast/id1643962763

Subscribe on Spotify: https://open.spotify.com/show/1FT9oqXSek3jMfiKrZPLQs

EPISODE CHAPTERS

0:00 - Introduction: one in four employers use AI, one in three job posts are fake

1:06 - Housekeeping and today's guest

2:34 - Kat's background and how job posts became the whole problem

4:34 - The number one mistake: you never defined the role

5:10 - What a job post is actually for

6:12 - The four categories every job post needs

7:48 - Hiring is a discipline, and the experience gap owners never close

9:07 - Even a hiring expert gets it wrong three out of four times

9:54 - Why using AI to recruit is a math problem

10:33 - AI cannot find what you cannot define

12:49 - Hiring for current revenue instead of future revenue

14:01 - AI is a tone thief: the battle of the LLMs

15:55 - Speed over results, and the screening checkbox that gets you sued

18:05 - Losing good candidates to your own calendar

19:12 - Publish your timeline before you post the job

21:14 - Why owners freeze on the final decision

22:02 - Define it up front so you know it when you see it

23:11 - Turning the job post into your interview scorecard

24:10 - What happens when you wing the interview

25:09 - Define what a good answer sounds like

26:42 - 37% quit in the first 90 days

27:42 - Using AI on interview questions means going line by line

27:56 - LLMs are not smart, they are word predictors

30:05 - Talk back to the tool

30:34 - The good enough culture that sunk hiring

31:30 - AI optimized resumes and the easy apply problem

32:32 - When 90% of your applicants look great

33:41 - Fake jobs, fake candidates, and the FBI investigation

36:08 - Corey's story: three candidates, three fake companies

37:38 - The deepfake test: turn your head to the left

38:21 - Indeed's agents and the monolithic algorithm problem

40:10 - Rejection loops that follow candidates across companies

41:35 - Actionable advice: spend your time on the job post

42:53 - Findability, and why made up job titles hurt you

43:48 - Results over speed

46:36 - Where to find Kat

47:25 - Corey's closing thoughts on how long hiring actually takes