Pay, ghosts, platform behaviour, remote work, visas, skills and inflated titles — measured over 7.4M postings, method and limits stated plainly.
Every argument about the job market is really an argument about measurement. One side says the openings are not real; the other says they are. One side says remote work is dying; the other says it was never counted properly. Nobody is lying. Almost everybody is guessing.
We do not guess. We crawl public job postings every day — 7,415,865 of them indexed, 3,624,739 with a full description — and we measure what the ads themselves say. This post collects the numbers we publish, the method behind them, and the parts we cannot answer. It is written to be quoted, so the figures are repeated where they matter.
One index, rebuilt daily from public sources — 7,415,865 postings indexed as of September 20, 2026. No sampling in the percentages below: they come from a full-corpus pass over the September 2026 edition, measured September 15, 2026 — every crawled record, no sampling.
Two samples appear in our work, and we always say which one applies:
The second one is the reason our ghost-job figures differ from survey numbers. A survey asks a person what they think happened. We look at what the source did.
Pay. Only 12.4% of postings state any pay figure. Of those that do, 7% give a min–max range, 11.1% of ranges are wider than $50,000, and 5.9% answer with "competitive" or DOE. The country gap is the louder story: 24.36% in the UK against 0.47% in Germany. The law and the practice are not the same thing. → /data/pay-transparency
Age and ghosts. Median listing age is 27 days; 29.8% of postings are older than 90 days. In the crawl archive the picture is sharper: 33.6% are older than 90 days, and 50.79% of those were still seen at the source within the last seven days. Old and still being served — 17.07% of everything we measured. The simplest explanation, that the links are broken, is dead: we probe employer URLs one at a time, and only 0.62% of 25,080 were dead. → /data/ghost-jobs · /data/ghost-index
Platforms. The same job behaves differently depending on where it is published. Workday postings median 12 days, lever postings 101 days, paycom 69 days. And 47.76% of Workday postings arrive without a publish date at all — we keep them, but they cannot be dated. On pay, ukg postings are the most transparent at 21.78%, against a market average of 12.4%. → /data/hiring-platforms
Remote. 3.1% of postings are remote-friendly — but only 5.2% state an arrangement at all, and among those, 60.2% are remote. The two numbers are not contradictory; they use different denominators, and both are true. Remote postings also average about 10% more. → /data/remote-work
Pay and longevity. Postings that state pay have a median age of 22 days; postings that stay silent, 27 days. Shorter on average, and we will not claim more than that. → /data/pay-vs-listing-age
Places. The same posting does not stay listed the same everywhere: 13 days in the UK, 53 days in India, 61 days in the UAE. → /data/listing-age-by-country
Experience. 2.8% of postings are marked explicitly entry-level; 12.9% state any years of experience, and 18.3% ask for a degree or licence. The seniority mix read from titles is flat — few juniors, few seniors, a very large middle. → /data/experience-and-credentials
Borders. 0.7% of postings mention visa sponsorship — roughly one in 140 — and 0.4% require existing work authorisation. Separately, 23.5% of postings are not written in English. → /data/working-across-borders
Skills and AI. "Communication" is named in 444,817 postings, ahead of leadership and customer service — soft skills outrank technical ones. AI/LLM tooling appears in 2.1%. → /data/skills-and-ai
Titles. Inflated titles are rare when you insist on evidence: 0.1% of postings. But "guru" appears in 958 postings and rank language like "lead" on roles without scope in 1,075. We would rather under-claim than produce a big, soft number. → /data/title-inflation
Structure. The market is atomised, not consolidated: 408,086 distinct employers, with the top 20 holding 6.3% (HHI 4.4). And 9.3% of postings appear on more than one board — which is why raw counts of "open jobs" always overstate the number of distinct roles. → /data/hiring-market
Intent. "Ghost job" carries an accusation. We can measure age and re-crawl activity; we cannot measure motive. A hard-to-fill requisition, a pipelined role, a slow approval chain and a listing somebody forgot to close all look identical from the outside. So we publish listing age and label it "still listed", never "open".
Whether the role is genuinely open. We see a page served today. We cannot see a decision made internally.
Why. Budgets, reorganisations, headcount freezes and quiet layoffs are invisible in a posting. Anyone who tells you the reason from the outside is inferring, including us.
This is why we publish the method next to the number, and why we would rather cut a figure than stretch it.
The measurement argument is turning into a legal one, and that changes what the numbers are worth.
In New York, bill S8877 passed the Senate and Assembly on 2 June 2026 and awaits the Governor's signature. It would require employers with 100 or more employees to say when a position is expected to be filled, and to remove a listing within two weeks of it being filled. Read carefully, that is a mandate for exactly the kind of measurement we already run: publish-date, expected-fill date, and how long a listing survives. If it becomes law, listing age stops being trivia and becomes a compliance signal.
In Europe, the EU Pay Transparency Directive moves pay disclosure from principle to obligation from June 2026 — which is the context for the gap between the UK's 24.36% and Germany's 0.47%.
And there is a gap in the middle: as a Congressional Research Service brief noted, no government statistics agency publishes a ghost-job count. Every number in circulation is either a survey (employers admitting to it) or a platform sample (one vendor's share of the market). We publish a crawl — the whole index, every day, with our method attached.
This is a beat, not a niche. The HR trade press reports on posting data constantly — HR Brew covers recruitment and retention numbers, HR Dive covers the AI shift in postings, SHRM analyses demand, HR Executive and Staffing Industry Analysts follow the disclosure bills. The business press reaches for job-posting data whenever official statistics feel late. Where we fit is narrow and useful: we are the crawl. When a reporter needs to know how long ads actually stay up, or how a platform behaves, or what share of postings mention a visa, we can answer with a number, a sample size and a method, from one index.
Every figure above has a page with the sample size attached: start at /data/key-figures, or take the raw files — /report/data.json, /report/data.csv — under CC BY 4.0. The method lives at /about-data, and what our corpus can and cannot vouch for is at /data-health.
Need a different cut? Every figure comes from one index, so we can extract almost any slice on request — a city, a role, an industry, a time window, or a comparison between platforms. Write to [email protected] and say what you need: we pull it, publish it with its method, and hand you the numbers. All we ask in return is a citation.
Cite as: DeUnemployment, "What 6.8 million job ads say — and the three things they cannot", https://deunemployment.com/blog/what-6-8-million-job-ads-say (September 2026 edition).
Our numbers below are ours; these are other people looking at the same question. We paraphrase and link, we do not quote.
Want a cut we have not published? A region, an industry, a role family, a time series, one country against another — we can pull it from the same corpus these posts are measured on.
Tell us what you are looking at: [email protected]. We answer either way.
Source & licence. All figures here come from our own crawl of public job postings, measured against the corpus published in the monthly report and its JSON (CC BY 4.0).
Cite as: DeUnemployment, “What 7.4 million job ads say — and the three things they cannot”, https://deunemployment.com/blog/what-6-8-million-job-ads-say (data as of 2026-09-20).