Ghost Jobs Are a Recruiter Problem, Not a Candidate One

Forty-three percent of job postings on major boards are either fake, stale, or have no actual hiring intent behind them. That number comes from a Clarify Capital survey of hiring managers themselves — not candidates whining on Reddit. The hiring managers admitted it.

And here’s the part nobody in our industry wants to say out loud: most of those ghost jobs weren’t posted by villains running fraud farms. They were posted by recruiters and TA leaders who genuinely meant to fill the role. They just got trapped in a workflow that punishes you for taking posts down and rewards you for leaving them up.

If you’ve ever left a req open “just in case a unicorn applies,” you’re part of the problem. I say that as someone who runs a recruiting platform and has watched the exact same pattern play out across thousands of accounts. Let’s actually talk about it.

What a ghost job really is

The candidate-facing definition of a ghost job is simple: a posting that has no realistic chance of being filled within a reasonable window. But that definition is too clean. In practice, ghost jobs fall into four buckets, and only one of them is genuinely malicious.

The first is pipeline farming — companies posting evergreen reqs to collect resumes for roles that don’t exist yet. This is the bucket the media loves, and it’s the smallest one in real life.

The second is the forgotten req — someone posted it three months ago, the hiring manager went quiet, the recruiter moved to another priority, and the post is still crawling through Indeed and LinkedIn collecting applications that nobody reads. Be honest. You have at least two of these right now.

The third is the political req — the role is “open” but the internal candidate already has it, or leadership hasn’t actually approved headcount, or the hiring manager is using the search to leverage a counter-offer for someone else. The req is theater.

The fourth, and the most common, is the dead-on-arrival req — posted with real intent, but the description was bad, the comp was off, the channel was wrong, and after the first 80 unqualified applicants the recruiter mentally gave up. The post stays live because taking it down feels like admitting defeat.

Three of those four are recruiter-side failures. Not candidate paranoia. Not LinkedIn’s fault. Ours.

Why this is hurting your actual pipeline

Here’s the uncomfortable downstream effect. Candidates have learned. Two years ago you could post a senior backend engineer role and get 300 applicants, maybe 30 worth a phone screen. Today you post the same role and you get 1,200 applicants, maybe 8 worth a phone screen. The signal-to-noise has collapsed because qualified candidates assume your post is fake and don’t bother, while everyone else mass-applies with the same boilerplate.

This is the part that should worry you more than the PR problem. Ghost jobs aren’t just bad for candidates — they’re poisoning the well you drink from. Every fake or stale post on the market makes the real ones less effective.

Greenhouse’s 2024 hiring data showed application-to-hire ratios roughly doubling year over year while time-to-fill barely moved. That’s not because candidates got worse. It’s because applying got cheaper (one-click apply, AI cover letters, auto-apply tools) and posting got lazier. The result is a market where neither side trusts the other, and recruiters are spending more time triaging garbage than actually evaluating humans.

The honest reason recruiters leave bad posts up

I want to push on this one because it’s the part the industry won’t admit. Why do reqs stay open after they’ve stopped working?

Because the alternative — taking a post down, telling your hiring manager the channel isn’t producing, and proposing a different approach — requires you to have a different approach. And most TA teams genuinely don’t. The job board is the muscle memory. Post it, wait, screen, repeat. If the posting isn’t working, the implicit assumption is that you wait longer or boost the post or rewrite the description. Almost nobody pulls the post and goes hunting.

That’s the actual root cause of ghost jobs. Not malice. Not laziness. A workflow that has exactly one tool in it.

If post-and-pray is your only motion, of course you leave the posts up. What else are you going to do?

The shift that actually fixes this

Here’s the thesis, and I’ll defend it: the era of posting a job and waiting is over for any role above entry-level. Not because posting doesn’t work at all, but because it shouldn’t be your primary motion anymore. It should be your supplemental motion.

The primary motion for any role where you can name the qualifications should be direct sourcing. And the reason you couldn’t do that before was bandwidth — you couldn’t reasonably hand-source 40 reqs as a recruiter. Now you can, because the search-and-outreach part is no longer manual.

This is why we built HiredGPT the way we did. You paste the job description (or just describe the role in plain English — “senior platform engineer in Austin with Kubernetes and Go, no FAANG resumes”) and it pulls qualified candidates from a pool of 850 million profiles. The query is the JD. There’s no Boolean string to build, no checkbox taxonomy to learn, no “AND (java OR j2ee) NOT junior” syntax that you have to remember from 2014. You describe the human you want, and you get a list of humans.

The point isn’t the tool. The point is that the bandwidth excuse is gone. If you can’t be bothered to source for a role, that’s now a choice, not a constraint.

What “kill your ghost jobs” actually looks like in practice

I’ll be specific because this is where most “rethink your hiring” blog posts go vague. Here’s a workflow that actually closes the ghost-job loop.

Audit your open reqs weekly. Not monthly. Weekly. Any post older than 21 days with fewer than three candidates moved to interview stage gets a decision: pull it, rewrite it, or escalate to the hiring manager about whether the role is real. You’d be amazed how many “open” reqs evaporate the moment you force a conversation about them. Your ATS view should be the source of truth here, not your memory.

Default to sourcing for anything mid-level or above. If you’re hiring a senior engineer, a regional sales lead, a controller — anything where the candidate has options — you should be reaching out, not waiting. Use candidate search or HiredGPT to build a shortlist of 40-60 qualified people in the first 48 hours. Then your job post becomes a landing page for that outreach, not a fishing net.

Automate the parts that are actually automatable. This is where Cortex earns its keep — you set the triage rules once (if candidate has X years in Y industry, advance; if location mismatch, auto-reject with a polite note; if there’s a referral tag, fast-track to hiring manager). The recruiter’s time stops going to sorting and starts going to conversations. That’s the only ratio that matters in this job.

Run outreach in batches, not one at a time. Manual InMails don’t scale and template blasts don’t convert. The middle path is a real sequenced campaign with personalization variables that actually work. Our campaigns dashboard handles the batching and the throttling so your domain doesn’t get flagged and your reply rates don’t crater. But the principle holds whatever tool you use: outreach is a campaign, not a one-off.

Close reqs that aren’t closing. This is the hardest one culturally. Recruiters are measured on fills, not on closures, so leaving a post open feels free. It isn’t. Every stale post is teaching candidates that your company posts fake jobs. The reputational cost is invisible and real.

The part where I’m honest about what AI doesn’t fix

I want to push back on my own pitch for a second, because the worst thing in this market right now is recruiters who think AI sourcing is going to do the entire job for them. It isn’t.

AI is excellent at the parts of recruiting that are pattern-matching at scale: finding profiles that match a description, drafting personalized first-touch messages, scoring resumes against a rubric, triaging an inbox of 1,200 applicants down to the 40 you should actually read. It is genuinely bad at the parts that are about human judgment: deciding whether a candidate’s career change is brave or flaky, sensing when a hiring manager is being unreasonable about comp, knowing which of two equally-qualified finalists is going to fit your team’s actual dynamic.

If you replace a thoughtful recruiter with an AI workflow, you’ll fill faster and hire worse. If you give a thoughtful recruiter an AI workflow, you’ll fill faster and hire better. The technology is a force multiplier, not a substitute. The recruiters winning right now understand this. The ones losing think one of two extreme things — that AI changes nothing, or that AI changes everything.

The actual call to action

I’ll keep this short because if you’ve read this far you don’t need to be sold.

If you’ve got a req that’s been open more than three weeks and you’re tired of sorting through the same 40 wrong applicants every Monday, stop posting and start sourcing. Post the role properly, but treat the post as your destination page — not your acquisition channel. Build a shortlist in HiredGPT, set up your triage rules in Cortex, and run a real outreach campaign instead of waiting.

Your dashboard will tell you within a week which motion is producing. I’d bet money it isn’t the post.

And if you want to see how the whole stack fits together before committing — the homepage walks through it without making you sign up.

Kill the ghost jobs. The market will be better for it, and so will your fill rate.