The hiring signal is broken. Here's what the data shows — and what to do next.

There is a concept in labor economics called costly signaling. The logic is elegant: if producing a high-quality job application — a tailored cover letter, a polished portfolio, a thoughtful answer to a screening question — requires genuine time and effort, then only candidates who actually have something to prove will bother doing it well. The effort itself is the signal. Employers trust it not because it is always accurate, but because faking it is expensive.

Generative AI has made faking it free.

This is not a metaphor. In the span of roughly two years, the marginal cost of producing a convincing, tailored application has collapsed toward zero for almost every candidate at every skill level.

The result, in economic terms, is a pooling equilibrium: a state where high-ability and low-ability candidates are indistinguishable from their written outputs. Everyone sounds good. The signal has lost its meaning.

The downstream effects are measurable. Simulations of AI-disrupted hiring markets show something counterintuitive: the technology that was supposed to democratize opportunity is, in its current form, doing the opposite. The best candidates are being hired less. The weakest are getting in more. And no one — not recruiters, not applicants — is getting useful feedback from the process.

Why recent graduates are most exposed

The groups hit hardest by this shift are those who relied most heavily on written signals to compensate for thin professional networks.

Recent graduates, career switchers, and international candidates built their credibility on the quality of their applications — precisely the credential that has been devalued. A first-generation college graduate with no alumni network and a strong cover letter now competes on equal written footing with someone who has neither her ability nor her effort. That is not progress.

Meanwhile, matching efficiency across the labor market has declined even as job vacancies remain historically elevated. Firms cannot find the right candidates not because the candidates do not exist, but because the filtering systems are overwhelmed. The noise-to-signal ratio has become untenable for manual review.

What still works — and why

The signals that survive are the ones AI cannot cheaply replicate. There are three categories worth understanding.

Process evidence over polished outputs

A finished portfolio piece tells an employer nothing about who made it. A GitHub commit history, a Figma version timeline, or a series of timestamped drafts tells a story of authorship that is nearly impossible to fake at scale. The insight here is uncomfortable: in a world where outputs are cheap, process has become the premium good.

Network referrals as a quality threshold

The data on this is not ambiguous. Referred candidates are dramatically more likely to be hired — not because employers are playing favorites, but because referrals serve as a pre-filtering mechanism that bypasses the noise entirely. The formal application portal is clogged. The informal channel is cleaner than it has been in a decade. The labor market is quietly reverting to reputation-based hiring, and most applicants have not noticed yet.

Live demonstration of capability

Employers are increasingly adding synchronous verification steps — pairing sessions, timed exercises, live coding, unscripted problem-solving — because they have learned that asynchronous submissions tell them little. If you are not preparing to perform under observation, you are preparing for the wrong test.

A note on what comes next

It would be easy to read this as a counsel of despair. It is not meant to be. The disruption of costly signaling is real, but it is also creating space for new signals to emerge — ones rooted in demonstrable skill, verified work history, and genuine human judgment. Skills-based hiring is not a trend; it is the structural response to a broken filtration system. Prompt engineering commands a 15% salary premium today not because it is mysterious, but because it is specific, verifiable, and hard to fake in an interview room. The question worth sitting with: what part of your work is genuinely yours — provably, demonstrably, irrefutably so? Start there.

---

Originally published on Medium on May 9, 2026.