I got tired of "remote" meaning "remote, US only", so I counted 2,798 companies

elon zuckerman — Sat, 18 Jul 2026 22:32:39 +0000

The itch

I don't live in the US. Every time I went looking for a remote job the same thing happened. A posting says "Remote," I get halfway through the application, and a location dropdown appears with fifty US states and nothing else. Or there's a line in the fine print: "Must be authorized to work in the United States." Remote, it kept turning out, meant remote inside one country.

After enough of those I stopped being annoyed and got curious. How bad is it actually? Is "worldwide remote" a real category or a rounding error? I couldn't find a number anywhere, so I built something that would give me one.

What I built

I run a remote job board that pulls postings straight from company applicant-tracking systems every day. Greenhouse, Lever, Ashby, and a pile of others. No aggregator sits in the middle, so a posting lands on my board within a day of going live, and I read the structured fields the company filled in rather than a secondhand summary. That detail matters for what comes next, because those structured fields turned out to lie sometimes.

Right now there are about 10,525 live postings on it. Once the feed was stable I could finally answer the question I actually cared about. If you're outside the US, how many of these can you actually apply to?

Half the work was throwing companies out

Before I could count worldwide roles I had to work out which "companies" were real employers at all.

I started with 2,798 companies that each had at least one open remote role. 402 of them, about 1 in 7, were clearly not real employers. Staffing agencies, recruiters, and gig platforms that present themselves as the employer. I dropped those. Another smaller set I couldn't confidently put on either side, so they stayed in a maybe pile and out of the final count. What was left was 2,378 companies I was sure about, with 7,483 open roles between them.

You cannot do this by name. Only 34 of those 402 had "staffing," "talent," or "recruiting" anywhere in the name. The rest look like normal companies until you read what they post. The tells are behavioral:

one "company" posting a nurse, a developer, an accountant, and a virtual assistant in the same week
"on behalf of our client" language sitting in the description
a website that sells hiring services instead of an actual product

None of that is a single regex, so I still eyeball the edge cases by hand. And this isn't a tech-only crowd, which surprised me. 813 of the real employers are hiring in engineering, but 639 are hiring in sales, 464 in customer support, and 447 in marketing.

The Cloudflare posting that broke my classifier

The hard part of this project was never the scraping. Greenhouse and Lever have predictable endpoints, Ashby makes you talk to a GraphQL API, and once pagination and a polite backoff are in place the fetching mostly runs itself. The hard part was deciding what a posting actually means, because the data is filthy.

My favorite example is a Cloudflare role. Its structured location field, the clean machine-readable one, was set literally to the string "Hybrid." My classifier read that, found nothing that looked like a country restriction, and fell back to its default, which was to file the job as open worldwide. That default was the real bug. A missing country lock is not proof that there is no country lock, and the classifier treated it as if it were.

The actual location was sitting several paragraphs down in the description body: "Available Locations: Austin, TX." I was only feeding the model the structured fields plus the opening stretch of the description, to keep token costs down, and that line sat well below the cutoff. So it never got read.

So a hybrid job in Austin, Texas got advertised on my board as something you could take from anywhere on earth. Exactly the failure I built the board to avoid.

The fix wasn't a smarter model. It was a dumb deterministic rule I should have had from day one. A hybrid or onsite job can never be worldwide, full stop, because "worldwide" only means anything for a fully-remote role. If the work mode isn't remote, the eligibility question is already settled, and no amount of buried free text can override it. One line of logic killed a whole class of false positives that no keyword matching would have caught cleanly.

The number, with the caveat attached

Of the 2,378 real employers, 216 have at least one role open to applicants anywhere in the world. That's 9%, roughly 1 in 11. The other 91% are region-locked, usually US-only, though not always. An EU-locked company still won't take an American but will happily hire outside the US, so region-locked is a wider bucket than US-only.

That 9% is generous by construction, since a company counts as worldwide if even one of its roles is. Count individual postings instead and you get 4.2%, which is 444 of all 10,525 live listings, a set that still includes the agency postings I pulled out of the employer analysis. The two aren't the same measurement, so I wouldn't stack one on the other. One is the share of employers with a global door open somewhere. The other is the share of individual listings you could actually apply to from abroad.

A few recognizable names do still hire worldwide, verified: GitLab, Remote (remote.com), Mozilla, Stripe. Most of the big remote hirers you'd expect turned out to be region-locked once I actually read their postings. ElevenLabs, ServiceNow, Grafana Labs, and Twilio all sit in that bucket right now.

One honest caveat. "Worldwide" is parsed from posting text, and the classifier still misses roles that describe eligibility in ways it doesn't catch. The Cloudflare fix cut false positives. It did nothing for false negatives. So treat 9% as a floor. The real share is higher, probably not by a huge margin, but higher.

If you want to poke at it

The board is free, no signup, no account, if you want to dig through the data yourself. https://globalremotely.com

What I keep chewing on is whether 9% is low because global remote hiring is genuinely hard, with payroll and tax compliance to sort out in every new country, or because companies default to "US-only" out of habit and legal caution when they could hire wider. If you've set up international hiring where you work, I'd like to know which wall you actually hit.