The Job You Posted Cannot Be Filled. Here's Why.

There is a recurring conversation inside almost every recruiting team. It usually happens around week eight of a search that should have closed by week four.

The hiring manager says the candidates aren't strong enough. The recruiter says the candidates don't exist. They pull up the job description together. They scroll through the requirements list. Somewhere between "10+ years of experience with technologies first released in 2021" and "expert-level fluency across four distinct domains, plus stakeholder management at the VP level," it becomes clear that the role, as written, cannot be filled.

Not by anyone. Not at any salary. Not in this market or the next one.

The role hasn't been failing because the talent pool is thin. The role has been failing because the document used to define it was written for a person who doesn't exist.

This is one of the most common failure modes in modern hiring, and one of the least examined. The job description is treated as the starting point of a search. In a significant share of cases, it is the reason the search never ends.

The purple squirrel is not a joke. It's a hiring strategy.

The recruiting industry has a name for the impossible candidate that job descriptions increasingly demand. Recruiters call them purple squirrels — mythical candidates who meet every single listed requirement, including the contradictory ones.

The term was a wry internal joke for years. It has stopped being funny because the requests have stopped being unusual.

Industry write-ups now routinely cite recruiters being briefed to find candidates like "an engineer with 30 years of AI experience" — for a discipline that, in its current commercial form, is barely a decade old. Postings asking for ten years of hands-on experience with frameworks that were released five years ago are common enough that the phrase "decade of experience with a five-year-old technology" has become recruiting shorthand for an unrealistic spec.

These are not isolated drafting errors. They are the visible symptoms of a structural problem in how job requirements get assembled.

Most enterprise job descriptions are produced by aggregation. A hiring manager lists what they wish the person could do. The previous role-holder's responsibilities get appended. HR adds compliance language. A peer team contributes a section because they'll also work with this person. Recruiting bolts on keywords the ATS is known to surface for. The final document is a layered composite of every stakeholder's hopes, fears, and unrelated requests, presented to candidates as a single coherent role.

The candidate looking at it has no way to know which requirements are real, which are negotiable, and which are vestiges of three previous re-orgs. The recruiter often doesn't either.

The cost of writing for nobody

LinkedIn's analysis of its own job-post performance data is unambiguous on what bloated requirements actually do.

Job descriptions under 300 words receive 8.4% more applications than the platform average. Responsibilities sections in the highest-performing posts run 9% shorter than those in lower-performing ones. The platform's own guidance to recruiters explicitly warns against the assumption that more detail produces better candidates — the data shows the opposite.

A separate LinkedIn data point puts the problem in candidate terms. Job seekers spend an average of 14.6 seconds reading the qualifications section of a job description before deciding whether to apply. The recruiter who spends three hours assembling a forty-bullet requirements list is producing a document that the people it's meant to attract will scan in less than a quarter of a minute, looking for two things: whether the role is plausible for them, and whether the compensation is mentioned.

If the requirements list reads as impossibly long or contradictory, the rational candidate doesn't apply. Not because they aren't qualified — because the document signals that the company doesn't actually know what it wants, or knows but expects to find someone it can't possibly afford. Either reading produces the same behaviour: the strong candidate moves on, and the role attracts the candidates who would apply to anything.

The recruiter then receives a pile of applications that don't match the requirements and concludes that the talent pool is weak. The talent pool isn't weak. The filter that produced this pile was designed to repel the people the role actually needed.

Forty-two percent of employers in 2025 said they have had to revise or rewrite their job descriptions because the original drafts attracted unqualified candidates. The standard interpretation is that the candidates were the problem. The harder interpretation — that the JD itself selected for the wrong pool — is the one the data actually supports.

What the AI-written JD has changed

The economics of producing a job description used to act as a soft brake on bloat.

Writing one took time. Each additional bullet on the requirements list was a small editorial decision that someone had to make. Stakeholders had to negotiate which line items mattered enough to include. The friction was inefficient, but it imposed a kind of discipline. Long, contradictory requirement lists were limited by the cost of writing them.

Generative AI removed the brake.

In 2025, creating job descriptions became the single most common use of ChatGPT inside recruiting workflows, accounting for roughly 30% of recruiter use of the tool. Approximately 70% of companies are projected to incorporate AI elements into their job postings during the same period. Internal LinkedIn surveys put the share of recruiters updating JDs to reference generative AI usage at 22% — a separate but related signal that the document is being rewritten faster than the underlying role is being rethought.

The downstream effect of AI-assisted JD writing is exactly what you would expect. The cost of producing a polished, comprehensive-looking document dropped to near zero. The volume of requirements that can be plausibly listed inside a single posting expanded. The hiring manager who would once have grumbled at writing a fifteen-bullet requirements list now has no friction stopping them from generating a forty-bullet one in twenty seconds.

The document gets longer. The role stays the same. The candidate pool gets thinner.

This is the AI-side mirror of what's happening on the candidate side. Candidates use AI to generate tailored applications for any posting in two minutes. Recruiters use AI to generate tailored postings for any role in two minutes. Neither side is producing a more accurate representation of what they have to offer. Both sides are producing more sophisticated-looking versions of the same noise.

The qualifications gap nobody actually measures

There is a widely cited claim, repeated for over a decade, that women apply for roles only when they meet 100% of the requirements while men apply at 60%. The claim shaped corporate diversity policy. It influenced how Sheryl Sandberg framed Lean In. It became received wisdom in recruiting.

It is also, as more careful research has now shown, substantially inaccurate.

A 2024 paper published in the European Journal of Social Psychology tested the original assertion using actual data. The real numbers: women apply when they perceive themselves as meeting around 56% of the requirements, men at around 52%. The difference exists, but it is roughly a thirtieth of what the popular version of the statistic claims.

The honest finding inside the research is more interesting than the myth it replaces. The gap that does exist is almost entirely driven by candidates' self-perception of fit, not by their actual qualifications. The same job ad produces slightly different reads of "am I qualified for this?" depending on who is reading it. And the longer and more requirement-dense the posting is, the more self-screening it produces — across every demographic, in every direction.

The implication for how requirements get written is direct. A job description is not a neutral list. It is a filter that disproportionately repels the candidates least likely to assume the requirements are negotiable — which, depending on the role and the market, can be the candidates the company most needs.

The companies adding more "must-have" requirements to their postings are not increasing the precision of their search. They are increasing the rate at which capable candidates self-disqualify before the recruiter ever sees them.

The credential layer underneath all of this

The bloated requirements list usually sits on top of a second, older layer of unnecessary spec: the degree.

The Burning Glass Institute's 2024 research, conducted with Harvard Business School, found that 43% of job postings requiring a bachelor's degree could effectively be performed by workers with alternative credentials or directly relevant experience. The degree wasn't filtering for capability. It was filtering for the historical convenience of the recruiter who didn't want to evaluate non-traditional backgrounds individually.

The skills-based hiring movement was meant to dismantle this layer. The announcements were real. The hiring data wasn't. The same research found that only around 3.5 percentage points of additional non-degree hiring resulted from a decade of public commitments to remove degree requirements — and that 45% of companies that pledged skills-based hiring fall into what the researchers called the "In Name Only" archetype, where the policy change produced no measurable hiring behaviour change.

So the average job posting in 2026 carries: a degree requirement that disqualifies most of the labour force for no defensible reason, plus a years-of-experience floor that often exceeds the age of the technology involved, plus a "must have" list inflated by AI generation, plus a "nice to have" list that hiring managers and ATS systems often treat as a second filter rather than a wish list.

Each of these layers, applied independently, would shrink the candidate pool. Stacked, they produce a posting that no real person matches.

What the unhireable JD does to the rest of the process

The downstream cost shows up across every metric a talent team is measured on.

Time-to-fill extends because the requirements as written cannot be met. The recruiter is forced into a quiet renegotiation with the hiring manager about which bullets are actually disqualifying, which happens in week six or week eight rather than week one — wasted weeks during which the role appears active but is functionally stuck.

Candidate ghosting increases because the candidates who do apply quickly realise the role they are being interviewed for doesn't match the role they applied to. They disengage. The recruiter logs it as a candidate-quality problem when it is actually a posting-fidelity problem.

Offer acceptance falls because the candidate who survives a multi-round process for a misrepresented role is the candidate who has had time to discover the misrepresentation. They counter-offer, or they ghost, or they accept and leave within a year — which surfaces six months later as a retention problem, not the hiring problem it actually is.

The recruiting team escalates by adding more screening, more rounds, more assessment — solving for the symptom by adding cost to the process. The hiring manager concludes the talent market is broken. The CFO concludes recruiting is inefficient. The candidate concludes the company doesn't know what it wants.

None of this is the talent market's fault. It is what happens when the source document at the start of the pipeline is written for a candidate that doesn't exist.

The exit isn't a better template. It's evidence about the real role.

The standard response to bad job descriptions is to write better ones. There are guides for this everywhere. Most of them recommend the same things: be specific, be concise, separate must-haves from nice-to-haves, lead with outcomes rather than tasks.

These are all reasonable suggestions. None of them addresses the underlying problem.

The underlying problem is that the job description, as currently produced, is a prediction document. It is the hiring manager's best guess at what the future role-holder will need to do, listed as requirements before the role has been performed. The further the prediction sits from the actual work that gets done, the more bloated and unrealistic the requirements list becomes — because the writer is hedging, adding contingencies, importing requirements from adjacent roles, and protecting themselves from the possibility that the chosen candidate turns out to be wrong.

The way to write a JD that maps to a real person is to anchor it in evidence about what the role actually requires. Not the wishlist. Not the composite of what every stakeholder thinks they want. The actual pattern of work that previous high-performers in similar roles demonstrated — the trajectory they brought into the role, the capabilities they applied once they were in it, the gaps they had on day one that turned out not to matter.

That requires looking at career patterns, not requirement lists. It requires asking what the people who succeed in a role like this have actually done — and writing the JD as a description of that signal, rather than a wishlist of credentials a system can keyword-match against.

The companies that close roles fastest in 2026 are not the ones with the most polished job descriptions. They are the ones whose descriptions reflect a job a real person has, in fact, done — and which therefore attract the candidates who can plausibly do it next.

The unhireable job description is a self-inflicted wound. The companies that recognise it as such are the ones that stop bleeding talent at the top of the funnel.

AgentR evaluates candidates against the patterns that actually predict success in a role — not the wishlist a job description was built around. If your roles are taking longer to close than they should and the candidates who do come through don't match what you wrote, the document is the problem. Let's talk.