28 AUGUST 2026 · 4 MIN READ

We checked 500 live job postings. 13 were already closed.

Everyone has a ghost jobs story. You apply, hear nothing, and later find the same role reposted. The usual claim is that employers list roles they have no intention of filling. That is very hard to prove from outside a company — so we measured the neighbouring thing, which is verifiable: how many job postings that are still listed have already stopped accepting applications.

What we did

We took a random sample of 500 postings from our index, all of them marked active, spread across the five biggest hiring systems. For each one we asked the employer's own system whether it was still open — Greenhouse, Lever and SmartRecruiters all publish an API that answers definitively. Where no API exists, we fetched the page and looked for the tell-tale redirects.

The check is deliberately biased toward saying "alive". A timeout, a bot wall, or a URL we could not parse all count as open. So the real number is at least this bad, and probably slightly worse.

What we found

13 of 500 — about one in 38 — had already closed. Eight returned a hard 404 from the employer's own API. Five were still being served happily by SmartRecruiters with an active: false flag: a page that looks completely normal in a browser and is not taking applications.

Why 2.6% is the flattering number

This index was rebuilt from scratch 24 hours before we ran the check. That is about as fresh as job data gets, and one in 38 was already stale.

Most places you search have not been rebuilt in the last day. Every day a listing sits unverified, the odds it is still real get worse — and nothing about the page tells you which way it went.

What we are not claiming

We did not measure ghost jobs in the strict sense. A closed posting is not evidence that the employer never meant to hire; roles fill, budgets freeze, and boards are slow to catch up. Proving intent from the outside is not something a crawler can do, and anyone selling you a percentage for it is guessing.

We also cannot tell you how the rate climbs as a posting ages. We looked. With only 13 dead postings, splitting them across age buckets produces numbers that bounce around with no pattern — that is noise, not a trend, and publishing it as a curve would be dishonest. Ask us again in a few months when the index has aged and there is a real series to plot.

The part that matters if you are job hunting

A dead posting costs you more than the ten minutes of typing. It costs the hour you spent tailoring a CV for it, and it quietly teaches you that applying does not work — when what actually happened is that you applied to a door that was already locked.

This is why Otto checks every posting against the employer's system before it reaches your shortlist, and again before filling in a form. It is unglamorous, it makes our job counts smaller than everyone else's, and it is the single most useful thing we do.

Method: 500 postings sampled at random from active listings on Greenhouse, Lever, SmartRecruiters, Workable and Ashby, checked 28 August 2026 against each system's public API where one exists, or by fetching the page. Ambiguous results counted as open. Sample size means the overall figure carries roughly a percentage point of slack either way, and the per-system rates should be read as rough rather than precise.