How long job postings stay up: we measured a 9-day median, not 48 hours

We apply to jobs for people, which means we reread employer career pages on a daily cycle and keep a copy of what was on them. That archive holds 761,645 job postings. 725,418 of those carry both a posting date and a last-seen date, and that pair is the entire measurement: the date the employer stamped on the listing, and the last day we found it still live on their own site.

Before the numbers, the limit that shapes all of them.

We have only been watching these pages from 16 July 2026 to 21 September 2026. That is 67 days. Everything below is bounded by it. The figures here were pulled on 20 September 2026.

The median posting stays up 9 days

Take only the postings that first appeared after we started watching, and that we then saw disappear. Both ends observed, nothing guessed at. That cohort is 73,535 postings.

The median time up is 9.0 days.

The same cohort, read as shares rather than percentiles:

• closed within 2 days: 16.6%
• closed within 7 days: 47.0%
• still up after 30 days: 10.8%

So slightly fewer than half of these postings were gone inside a week, and slightly more than half were not.

Where we disagree, and with whom

Two pieces of received wisdom sit on either side of our number, and we disagree with both, in opposite directions.

The first is the 48-hour rule. Apply inside two days or do not bother, usually stated as though most roles are effectively decided by then. Our measurement does not support that. 16.6% of the postings we watched closed inside two days, and 47.0% closed inside seven, so most of the cohort was still live a week after going up. A posting from last Tuesday is sitting near the middle of the distribution, not off the end of it.

The second runs the other way. Indeed's career-advice page on this exact question says "Most job postings stay active for 30 days", a page last updated 15 June 2026 and read by us on 20 September 2026. We measure much shorter than that. Only 10.8% of our clean cohort was still up at the 30-day mark, so "most" is nowhere near what we see.

Now the part where the gap is partly our fault. Our window is 67 days long and it hides slow searches by construction, which the last section works through, so the real median is above 9.0 days. We doubt it is 30. Indeed's page gives no sample size, no window and no method next to that figure, so there is nothing there to reconcile against, and between an unsourced round number and 73,535 observed postings we will use ours and keep saying out loud that it reads short.

The concession the 48-hour crowd has earned: the 25th percentile is 3.9 days. One posting in four was gone before the fourth day, and nothing on the page tells you which quarter you are looking at. Applying early costs nothing beyond time you were going to spend anyway. The argument here is not against speed, it is against the other half of the rule, the half that says to skip anything older than a weekend.

Two other cohorts, both of which run longer

Widen from the clean cohort to every posting we have ever seen disappear, including ones that were already up before we started watching, and the numbers move up. That group is 87,581 postings. Median 11.9 days. The 75th percentile is 29.9 days, the 90th is 50.8 days, and the 99th is 237.8 days. The longest-lived posting in that set had been listed for 1,814 days.

Then there is the far bigger group: 624,515 postings that are still live right now.

Their median age so far is 18.4 days. 27.3% have already been up more than 30 days, 10.8% more than 60 days, and 4.9% more than 90. The oldest single posting we can see has been listed for 4,159 days.

Every one of those is a lower bound. Those postings have not closed, so their real lifespan is longer than the age we can see, and we will not know by how much until they do close. The honest summary is that the long tail in our corpus runs much longer than a week, and 67 days of watching only shows us the start of it.

What this changes for a job hunt

Stop treating posting age as a filter. If your search is pinned to "posted in the last 24 hours" you are looking at a thin slice of what is open, and you are throwing away a large share of listings that will still be up a fortnight from now. A posting two weeks old sits past our median of 9.0 days but well inside the 75th percentile of 21.6. It is not dead.

The practical version:

• Apply early when you can. A quarter of the cohort was gone by day 3.9, and early costs you nothing.
• Do not skip a posting because it is old. At 14 days it is past our median and still inside the middle half of the distribution. At 30 days, 10.8% of the cohort we watched was still up.
• Treat very old listings with suspicion, not certainty. Some of the long tail is a genuinely slow search. Some of it is a listing that nobody ever took down, and from the outside those look identical.
• Spend the saved time on volume and on fit, not on refreshing a feed. Age is a weak signal. What the posting actually asks for is a strong one, and that is worth reading before you answer a single question.

If you want to browse by employer rather than by date, our company pages list what each one currently has open. For what the market looked like in aggregate a month before this measurement, we counted roles and skills across the corpus in the August 2026 tech job market report. Most of these listings are managed by an applicant tracking system, which is also what decides how your application is parsed once you send it, and the formatting side of that is covered in our résumé notes for 2026.

How we counted, and what it cannot tell you

What was counted: every posting we hold that carries both a posting date and a last-seen date, 725,418 of them out of 761,645 total. Postings missing either date are excluded, because you cannot measure a duration with one end of it. "Closed" here means the last day we found the posting live on the employer's page. "Up" means the number of days between the two dates.

The clean cohort of 73,535 is the subset where both events happened inside our observation window. It is the strictest set we can build. The headline comes from it.

Two things about that headline are weaker than they look, and both cut against us.

The 9.0 day median is biased short. A posting that outlives our 67-day window is still live today, so it never enters the cohort at all. Conditioning on "we watched it close" over a short window systematically over-samples short-lived postings, and drops precisely the slow searches that would pull the median up. The true median is longer than 9 days. We cannot say by how much from 67 days of watching, and we are not going to estimate it.

Anything that looks like a hard ceiling near 62 days inside that cohort is an artifact of the window, not a fact about hiring. It is the arithmetic of when we started looking. We are naming it here so nobody, us included, reads it back later as a finding about how long employers keep a role open.

Two more limits worth stating. A posting disappearing does not mean the role was filled, only that the page stopped serving it, and we cannot tell those apart from outside. And the employer's own posting date is whatever they put on the page, which is sometimes the date of a repost rather than the date the search began.

We will rerun this in a few months. The window will be longer by then, the clean cohort will be bigger, and the median will almost certainly come out above 9.0 days. When it does, the new number replaces this one on this page.

Until then the habit worth building is small. Check the posting date, note it, apply anyway unless the thing is months old.