How fast do job postings actually disappear?

How fast do postings actually come down?

In a single observation window on July 31, 2026, we recorded 3,201 postings that employers took down from their own hiring systems. Across the surrounding 29 hours, 6,436 postings came off the board in total. We aged each confirmed takedown against its original posting date, then combined those with the age mix of the quarter million postings live on our board (246,599 at measurement) to estimate a survival curve: the share of postings still up, by days since they went live.

The shape is steep at the front. The riskiest day in a posting’s life is its first: about one in five brand-new postings is gone within a day. Of everything employers took down, 35 percent had been up for less than three days, 46 percent less than a week, and 90 percent less than two weeks. The median taken-down posting lasted about seven and a half days, and by day 13 less than half of a cohort of postings is still live.

Why do postings vanish so quickly?

Three ordinary reasons, no conspiracy required. Roles genuinely fill, and the fastest fills skew toward postings that attracted strong early applicants. Openings get paused or pulled when budgets and plans change, which happens on the employer’s clock, not yours. And some postings come down only to go straight back up, so the listing you see today may be a rerun.

That last one is measurable. One in eight postings live on our board right now carries a repost label, meaning we detected the same role coming down and returning, and that number is a floor, not the true rate, because our repost detection covers only part of the board so far. A reposted role is not necessarily fake, but its “posted yesterday” date is not telling you what you think it is.

What does this mean for when you apply?

The folk wisdom in every job-search forum says apply within the first day or two. Our data says the folk wisdom is right, for a mechanical reason: the posting itself may not survive the week, and the takedowns cluster hard in the first few days.

Practically:

• Treat the first 72 hours after a posting goes live as the real window, and prioritize fresh postings over backlog
• Check the actual posted date before you invest in a tailored application. If a site does not show one, assume the worst
• Watch for repost labels. A rerun posting deserves a more skeptical read of its date
• An older posting still live is not automatically dead, some teams run long processes, but do not build your week around page five of a search

What is the ethical way to collect job data?

A measurement like this is only as honest as the data behind it, so it is fair to ask how we gather ours. Our rules are simple, and we think they are the only defensible way to run a job platform.

• Read the employer’s own public hiring system, never another job site’s listings. Copying aggregators multiplies their staleness and their scams
• Public pages only. No login walls, no personal data, and request rates a careers site would never notice
• Honor reality immediately: when a posting disappears at the source, it disappears from us. We do not keep dead listings around to look bigger
• Label what we know: real posted dates, and reposts marked as reposts
• Expire honestly: nothing on our board is older than 30 days, ever

Continuous is the operative word. On an ordinary day our pipeline re-confirms more than 200,000 individual postings, one by one, against the employer systems they live on (210,213 in the day we measured), and the nightly expiry sweep touches every posting on the board on top of that (251,609 that night). Call it nearly half a million posting-level checks a day, plus a fingerprint check of roughly 25,000 employer boards every 15 minutes. None of this is complicated. It is just expensive to do continuously, which is why most sites do not. But it is the difference between a job board you have to double-check and one you can act on.

How did we measure this, and what are the limits?

The takedown sample comes from one observation window on July 31, 2026: 3,201 postings confirmed removed at the employer’s own system, each aged against its posting date. The population behind it is larger than any single day’s board: since late June we have followed 383,140 individual postings through their life cycles, of which about a quarter million are live at any moment. Daily removal rates were estimated by scaling that window to a full day and dividing removals at each age by the number of live postings of that age, then chaining those rates into the survival curve above.

Three honest limits. First, this is one day’s snapshot, and we will keep re-measuring; treat the exact figures as first estimates, not constants. Second, because our board expires postings at 30 days, the curve describes the fresh side of the market and says nothing about postings that sit for months, so it cannot measure the long-lived “ghost job” phenomenon regulators are now examining; if anything, our numbers understate how strange the old tail is. Third, the repost share is a floor, since detection covers only part of the board.