Your Job Description Is a Self-Defeating Filter. AI Wrote It and Nobody Questioned It.

AI-written job descriptions are now standard practice — LinkedIn reported that over 60% of job postings at mid-to-large employers were partially or fully drafted by AI tools in 2025 [1]. The problem isn't that the AI writes poorly. The problem is that it writes confidently, and nobody reads it critically before it goes live.

How a JD Goes From "We Need Someone Good" to "Unicorn Only"

When a hiring manager pastes their last job description into an AI tool and asks for an updated version, the AI does something predictable: it over-indexes on what's already there and adds elaboration. A "nice to have" in the original becomes a "required" in the revision. A general "communication skills" bullet becomes "demonstrated ability to communicate complex technical concepts to C-suite stakeholders in written and verbal formats." The signal dilutes; the requirement count climbs.

The average job description now lists 15–20 required skills [2]. Most of those roles historically hired people with 7 or 8. The gap isn't because standards rose. It's because AI expanded the template without anyone asking whether the additions were real.

Candidates, for their part, have learned to treat JDs as aspirational wish lists rather than actual requirements. But the ones who take requirement lists at face value — early-career candidates, career-changers, people returning to work after a gap — self-screen out before a recruiter ever sees them. You never get to count what you lost.

The Talent Shortage That Isn't

There's a persistent narrative that the talent pool is thin. In specialized fields, that's real. But a meaningful percentage of "talent shortage" diagnoses are actually JD problems in disguise.

The research is consistent: qualified women apply to jobs when they meet roughly 100% of listed requirements; qualified men apply when they meet around 60% [3]. If your JD lists 18 requirements and a strong candidate meets 12, half your potential applicant pool reads that gap as a reason to skip. You have introduced a demographic filter you didn't intend, and your ATS never flagged it because everyone who did apply looked fine.

The more insidious version: requirements that made sense for one iteration of a role get inherited by AI across every future version. A head of engineering who wanted cloud certification because of a specific architecture decision three years ago gets that criterion baked in permanently. The role evolves; the JD doesn't, because the AI is drawing on the last document you gave it.

What a Good JD Actually Filters For

The purpose of a job description is not to enumerate every skill a useful employee might have. It is to answer two questions a strong candidate is asking: Is this worth my time? and Can I actually do this?

A JD that answers those questions clearly and honestly creates a tighter, more relevant applicant pool — not a larger one, but a more qualified one. That means separating hard requirements from genuine preferences. Writing specifically about the actual work, not generic role archetypes. Being honest about level: a "senior" title that pays junior rates will attract candidates who either don't read carefully or have no other options. Neither is who you're trying to hire.

The AI, left to its own devices, will produce something that sounds comprehensive. It will not tell you that the "5 years of Salesforce experience" requirement you just included will eliminate most candidates from a market where the median experienced hire has 3. That judgment has to come from a human who actually knows what the role needs, not what the template suggests.

How Asendia AI Fixes the Problem at the Right Stage

If you've published a JD that over-specced the requirements, there's a second problem waiting downstream: the candidates who do apply get screened against an equally rigid checklist that penalizes people who don't tick every box. One bad document compounds into a bad pipeline.

Asendia AI is a voice-first AI recruiter that screens candidates 24/7 — but the screening logic is configured by your team against what the role actually needs, not what the JD happens to say. A hiring manager can specify that cloud certification is preferred but not required; Asendia's structured screening conversation will probe for equivalent experience without eliminating someone who has it in practice but not on paper. Every applicant gets a real conversation, at any hour, within hours of applying.

What comes back into your ATS isn't a pass/fail binary on 18 requirements. It's a qualification summary and candidate quotes that give your recruiter actual signal about fit — including edge cases where a candidate doesn't formally hit the checklist but clearly has the capability. Recruiting agencies use Asendia to absorb high-volume campaigns without letting automated screening amplify bad JD logic at scale. The platform connects directly to your existing ATS with no parallel workflow to manage. If you've been thinking about whether your ATS is even set up to handle this kind of signal, this post on what your ATS is actually built for covers why most teams are using it wrong from the start.

The compounding issue — AI generating JDs that feed AI screening that filters by AI-generated criteria — is one of the more underreported dynamics in hiring right now. Breaking that loop requires human judgment at the JD stage and screening tools flexible enough to interrogate capability rather than enforce checklists.

Final Word

The talent shortage is real in some markets, for some roles. In others, you manufactured it at the keyboard. AI-generated job descriptions that inflate requirements, inherit legacy criteria, and use template logic where judgment is needed are quietly creating a top-of-funnel problem that looks like a supply problem. The candidates who would have been strong hires are filtering themselves out before you ever count them. The fix isn't better AI for writing JDs — it's a human who knows what the role actually needs and is willing to delete requirements that are aspirational rather than essential. That review takes about 20 minutes per role and will do more for your pipeline quality than any downstream optimization you're currently running.

Ready to transform your hiring strategy? Schedule a Demo with our founders today!