Only a Third of "Remote" Jobs Are Remote — OmniGrade Blog<br>You find it. The role is perfect. The stack is what you want to be writing, the<br>team looks sharp, the mission is one you’d actually get out of bed for, and it says<br>remote, so you can keep the life you’ve built around not commuting. You click. You<br>start filling it out, already picturing the first day. Then, somewhere near the<br>bottom, the fine print: “hybrid, three days a week in our San Francisco office.”<br>You didn’t miss it. It wasn’t in the part you filtered on. And the letdown isn’t<br>the twenty minutes you spent on the form. It’s that you let yourself want it.
That’s the real cost, and it’s why this stings more than it should. Everyone who<br>has looked for an engineering job in the last two years has spent that emotional<br>bandwidth on a role that was never what it advertised. What nobody had was a<br>number. We have one now.
OmniGrade reads the full body of every posting it ingests and works out how the<br>job is actually done, remote or hybrid or onsite, from what the posting says<br>rather than from the label on it. That lets us do the thing a job board can’t:<br>take a role the platform swears is remote and check it against the words in the<br>posting. So we did, across every listing hosted on Ashby, one of the<br>applicant-tracking systems a large slice of startups run their careers pages on.
The tag is wrong two-thirds of the time
In July 2026, of 17,191 engineering roles Ashby tagged remote, 62.9% are<br>hybrid or onsite once you read the body. 37.1% are actually remote. The tag is<br>right a little more than a third of the time. On the other two thirds it’s an ad.<br>The most common shape is mundane: a role tagged remote on a posting whose own<br>location line reads something like “Hybrid — San Francisco, CA.” The label and the<br>fine print sit one inch apart and disagree, and the label is the one you filtered<br>on.
The five species of “remote”
Fake remote isn’t one thing. Sort those 17,191 postings by what the body actually<br>commits to and they split into five kinds:
What “remote” turned out to mean<br>Count<br>Share
Hybrid — N days in office, or anchored to a metro<br>10,670<br>62.1%
Geo-fenced — country- or region-locked, mostly US-only<br>4,496<br>26.2%
Genuinely remote — work from anywhere<br>1,850<br>10.8%
Onsite — remote in the tag only<br>175<br>1.0%
Read the middle rows. Genuinely-work-from-anywhere is 10.8% . About one in nine<br>postings sold as remote is the thing the word is supposed to mean. The other<br>eight in nine attach a string: come into the office some days, or live in the<br>right country, or, for a small hard core, just show up.
Who fakes it most
If this were sloppiness it would be random across companies. It isn’t. Sort the<br>same postings by the company’s funding stage and a gradient falls out:
Stage<br>Postings<br>Fake-remote rate
Bootstrapped<br>458<br>35.4%
Seed<br>995<br>47.3%
Series C<br>2,021<br>50.9%
Series A<br>1,708<br>61.9%
Series B<br>2,172<br>65.3%
Public<br>363<br>70.5%
Acquired<br>660<br>73.2%
Series E+<br>2,318<br>76.6%
The companies most honest about remote are the ones that have the least of it.<br>Bootstrapped and seed-stage sit at 35% and 47%. The worst are the ones with the<br>most of everything: public, acquired, and late-stage rounds, all north of 70%,<br>topping out at 76.6% for Series E and beyond. It isn’t perfectly monotonic, Series<br>C sits oddly low, but the two ends aren’t ambiguous. A late-stage company is<br>roughly twice as likely to launder office attendance as remote as a bootstrapped<br>one.
Seniority doesn’t buy you out of it either. The rate barely moves across levels,<br>62 to 65%, until the very bottom, where junior roles hit 86%. If you’re<br>early-career and filtering for remote, roughly seven of every eight matches are<br>going to want you in a building.
The lie is wrong at the source, so everyone repeats it
Here’s why a wrong tag matters more than one letdown. The company’s own<br>careers page is the source of record. LinkedIn, Indeed, the aggregators, the<br>niche remote boards, none of them re-reports the job. They ingest that feed and<br>pass the tag straight through, and none of them reads the body either. So when the<br>source says remote and means hybrid, nothing downstream catches it. The error gets<br>copied, onto every board, under the same “Remote” filter, all of them faithfully<br>repeating a claim that was already wrong where it started. You can switch job<br>boards all you like. You are filtering on the same broken field.
This is Ashby, one applicant-tracking system among several, and the shape it shows<br>almost certainly isn’t unique to it. We’ll take the same read to the other big<br>platforms in a later post. But the mechanism travels no matter where the posting<br>starts: the tag is set where the job is born, everything downstream trusts it, and<br>the only place the truth lives is a body that nobody reads at scale.
Why it happens
Reading this as carelessness gives it too much credit. A remote tag is a funnel<br>widener. Switch it on and your posting reaches every engineer running a remote<br>filter, nationwide or...