We Said 58% of Jobs Were Remote. That Was Wrong.

Bars scaled to 100%

Baseline is the 58.6% naive figure. Most real sources sit well under it.

Remote % = postings with is_remote=true / all active postings from that source. Himalayas, Jobicy, and Remotive are remote-only boards by design, every posting they carry is remote, so their volume mechanically sets a floor under the pooled rate. Sources under 3,000 active postings are omitted from the table but included in every pooled total.

Source: Seeker job corpus, 363,775 active job postings, snapshot 2026-09-04. seekerscore.com/insights/how-remote-is-the-job-market-2026

The finding

Our corpus holds 363,775 active postings from 20 sources. 213,215 of them, 58.6%, are tagged remote. But this is a pooled count, and pooled counts are only as representative as the mix of sources feeding them. Himalayas is a job board that exists specifically to list remote roles; every one of its 158,116 postings in our corpus is remote, and that single source is 43% of everything we hold. Take it out and the pooled rate drops to 26.8%. Take out the other two remote-only boards in our mix, Jobicy and Remotive, another 1,554 postings, and it settles at 26.2%. The correction is not fragile to which remote-only board you remove first, dropping Himalayas alone already gets you 96% of the way to the final number.

How we measured it (and what we did not claim)

We queried the same production jobs table every Insights page reads, filtered to active = true, and grouped by source and the is_remote column (a boolean, our data has no hybrid distinction). Twenty sources feed the corpus; we grouped every source with fewer than 3,000 active postings into the pooled totals without a separate row, since a source with a few hundred postings cannot move a corpus-wide percentage on its own. This number describes the composition of our corpus, not a verified fact about the U.S. labor market. Our source mix is a product of which job boards we chose to ingest, not a random sample of every job posting that exists, so a genuinely representative remote share would need a sampling frame no job aggregator, including us, actually has. What we can say with confidence is narrower and still useful: whatever remote-rate figure you see quoted from an aggregated jobs dataset is only as trustworthy as that dataset's source mix, and a single heavily-weighted, structurally-biased source can double a headline number.

Why one board can move a market-wide number

This isn't a Himalayas-specific problem, it's what happens whenever one source in a pooled count is both large and systematically different from the rest. A board that only lists remote roles doesn't need to misrepresent anything to distort an aggregate; it just needs enough volume. Ours happened to be large enough (43% of the corpus) that removing it more than halved the headline figure. Any dataset built by combining job boards, ours included, inherits this risk whenever one source's ingestion volume and its remote-listing rate both sit far from the rest.

Frequently asked questions

Related Data