Two Clocks Run on Every Job Posting, and They Disagree About What Is New
Based on 257,534+ analyzed job listings · Updated 2026-08-16
Every job posting carries two dates. There is the date the board first saw it, which is what almost every freshness figure on the internet is computed from, including ours. And there is the date the employer says they opened the role, which the feeds hand us and nobody reads. We compared them on 49,997 active postings that carry both. They agree almost exactly about what is old: 54.2% of postings are inside 30 days on the employer's clock and 53.8% on ours. They disagree completely about what is new. On our clock 23.2% of live postings are less than a week old. On the employer's, 10.3% are. Look at a single day and the gap is 34x: 6.8% of postings entered our index in the last 24 hours, and 0.2% were opened by the employer in that time. The entire discrepancy sits in the bucket everybody cares about.
28 days
Median age when you first see it
Measured on the employer's own posted date, across 49,997 postings that carry one
23.2%
Less than a week old, board clock
Share whose first sighting by our index was inside 7 days
10.3%
Less than a week old, employer clock
The same postings, dated by when the employer says the role opened
0.2%
Posted in the last 24 hours
Against 6.8% that entered our index in the last 24 hours, a 34x gap
54.2% vs 53.8%
Where the two clocks agree
Share inside 30 days, employer clock and board clock. The disagreement is only at the fresh end
4.8%
Older than 90 days
Still active, still served, opened more than three months ago
Share of live postings inside each age band, by which clock you use
Same 49,997 postings measured twice. The two clocks converge as the window widens: at 30 days they are within half a point of each other. Everything they disagree about is inside the first week.
Median posting age by source, on both clocks
| Source | Postings compared | Median age, employer clock | Median age, our clock | Rows where the feed moved the date forward |
|---|---|---|---|---|
| himalayas | 23,179 | 33 days | 27 days | 26% |
| workday | 9,637 | 24 days | 25 days | 9% |
| amazon | 3,075 | 27 days | 25 days | 12% |
| oracle_cloud | 3,060 | 30 days | 34 days | 14% |
| greenhouse | 2,992 | 11 days | 38 days | 71% |
| adzuna | 2,219 | 30 days | 15 days | 5% |
| 1,316 | 24 days | 31 days | 24% | |
| apple | 1,240 | 81 days | 48 days | 11% |
| usajobs | 799 | 23 days | 15 days | 6% |
| kaiser | 537 | 23 days | 23 days | 18% |
| citi | 533 | 12 days | 0 days | 14% |
| lever | 510 | 163 days | 43 days | 2% |
| All sources pooled | 49,997 | 28 days | 26 days | 21.2% |
Every source with at least 500 comparable rows is shown; the seven smaller ones are absent from the table but included in the pooled row. The last column is the share of rows where the employer date is LATER than our first sighting, which means the feed restated the date after we had already seen the posting. greenhouse does this on 71% of rows, which is why its employer-clock median reads younger than its board-clock median rather than older.
Source: Seeker job corpus, 257,534 active job postings, snapshot 2026-08-16. seekerscore.com/insights/job-posting-age-two-clocks-2026
The finding
A job posting is a median of 28 days old the first time it can appear in front of you. That is not a claim about our ingestion lag, because our own clock gives a nearly identical median of 26 days. It is a claim about what is in the pool: the live market is dominated by roles that have been open for weeks, and the small slice that genuinely opened this week is about half the size a board's own date makes it look.
Why the two clocks diverge only at the fresh end
A board's first-sighting date can never be earlier than the day it started reading a feed, and it resets whenever a posting enters the index for the first time. So a role that has been open for three weeks and reaches a new board today is dated today by that board and three weeks ago by the employer. Once you widen the window past a month, both clocks have had time to see the same postings and they converge: 54.2% against 53.8%. Inside a week they cannot converge, because one of them is measuring the employer's decision and the other is measuring our discovery of it.
What this changes about applying early
The common advice is to apply within the first few days, on the theory that early applicants get read. If you are using a board's posted date to decide what counts as early, you are being told a role is new when it is a median of four weeks old. On the employer's clock only one posting in ten is under a week old and one in five hundred is under a day. The practical version of the advice is that speed is worth much less than it looks, and the postings where it genuinely applies are rarer than any board's freshness figure implies.
The dates we do not trust, stated plainly
Two things push against this finding and both are in the table. First, 21.2% of rows carry an employer date LATER than our first sighting, which is a feed restating the date on a posting we already had. greenhouse does this on 71% of its rows and its employer-clock median therefore reads 11 days against 38 days on ours. Every one of those rows makes the pooled result look YOUNGER than the truth, so 28 days is a floor rather than a midpoint. Second, a handful of employer dates are not credible: lever's oldest active posting claims 2014-01-23, and 4.8% of all rows claim to be more than 90 days old. We have not deleted those rows, because deciding which employer dates to disbelieve is exactly the kind of judgement that turns a measurement into an argument. Both effects are disclosed rather than corrected.
How we measured it (and what we did not claim)
Population: a 20% systematic sample of the 257,534 active postings in the corpus on 2026-08-16, of which 49,997 carry an employer-supplied date that parses as a calendar date. Both clocks are computed on those same 49,997 rows, so the comparison is within-posting rather than between two different populations. The employer date is read from the raw feed payload and is never written by us. Validation that it is genuinely the employer's date and not our own echoed back: the earliest values run to 2014, 2016 and 2019, years outside our corpus window entirely, which begins 2026-02-07; and on the sources where a feed publishes a real date the employer date precedes our first sighting on 90% to 100% of rows (ashby 100%, usajobs 95%, adzuna 92%, findwork 91%, lever 90%). Coverage is not universal: 17 of 19 sources publish the field on essentially every row, kaiser publishes it on 75%, and reed publishes it on none, so reed is absent from this analysis entirely. We are NOT claiming a time-to-fill, a time-to-close, or that older postings are dead. A posting that has been open 60 days may be actively hiring. The only claim here is about how old the pool is and how much of the apparent freshness on any job board is the board's own discovery date.
Frequently asked questions
How old is the average job posting when I see it?
A median of 28 days, measured on the employer's own posted date across 49,997 active postings that publish one. Only 10.3% are less than a week old and 0.2% were opened in the last 24 hours.
Why do job boards say a posting is new when it is not?
Most boards date a posting by when they first saw it, not when the employer opened it. A role that has been open three weeks and reaches a new board today gets dated today. On our own index 23.2% of live postings look less than a week old, against 10.3% on the employer's clock.
Does applying early to a job posting actually matter?
Less than the advice implies, because far fewer postings are genuinely early than the dates suggest. One posting in ten is under a week old on the employer's clock and roughly one in five hundred is under a day. The pool you apply into is mostly roles that have been open for weeks.
How many active job postings are more than three months old?
4.8% of postings that carry an employer date claim to have been opened more than 90 days ago and are still listed as active. Some of those dates are not credible, which we state on the page rather than quietly filtering out.
See where you stand
Upload your resume to see your skill gaps, match scores, and which roles fit your background.
Analyze my resumeRelated Data
We Analyzed 1,835 “Entry-Level” Job Postings. Most Still Demand Experience.
“Entry-level” / junior postings analyzed: 1,835
The Highest-Paid Skills in 2026, From 73,374 Job Postings That Disclosed Pay
Job postings we analyzed: 73,374
Backend Engineer Market: What 200,000 Job Listings Tell Us
Active backend roles: 24,600+
The 25 Fastest-Growing Skills in 2026 Job Postings
Jobs analyzed: 133,000+
Guides
Dataset & citation
- Dataset
- 257,534+ active job postings
- Collected
- 20% systematic sample of active postings, snapshot pulled 2026-08-16
- Last verified
- 2026-08-16
- Methodology
- How we measure
Suggested citation
SeekerScore. "A Job Posting Is Already 28 Days Old the First Time You See It (2026 Data)." Analysis of 257,534+ job postings, 2026-08-16. https://www.seekerscore.com/insights/job-posting-age-two-clocks-2026
Data derived from Seeker's job corpus of 257,534+ listings across multiple sources. Updated 2026-08-16. Individual results vary based on resume content, target market, and role specifics.