The door, measured
Handshake counted the share of full-time job postings on its platform that offer visa sponsorship. In 2023 it was 10.9%. In 2026 it is 2.6%. That is roughly one in nine falling to roughly one in thirty-eight, in three years, and the steepest fall is in tech — the sector that spent a decade telling international students it was the way in.
Read the unit carefully, because it decides what the number can carry. This is a share of postings, on one platform. It is not the share of employers who sponsor, it is not the number of sponsored jobs that exist, and a platform whose posting mix shifts will move this figure without a single employer changing policy. What it does establish is a direction and a magnitude, and both are large.
| Year | Postings offering sponsorship |
|---|---|
| 2023 | 10.9% |
| 2026 | 2.6% |
A second number, from a different instrument
Separately, a survey of American companies asked whether they were open to hiring foreign business-school graduates. 29% said yes in 2026, down from 33% a year earlier and 55% in 2022. Bloomberg reported it at the end of June.
That is a survey of stated willingness. The Handshake figure is a count of what was actually advertised. They are not interchangeable, and an edition that averaged them, or presented one as confirming the other, would be inventing a precision neither has. What is fair to say is narrower and still striking: one instrument watching behaviour and one instrument asking intent have moved in the same direction over the same period.
Note the shape of the survey series too. The fall from 55% to 33% took four years; the fall from 33% to 29% took one. Whether that is an acceleration or a floor being approached is not something a three-point series can settle, and we are not going to pretend otherwise.
The number nobody puts in a headline
Underneath the policy story sits an arithmetic one. International students apply to twice as many jobs as their domestic peers and receive thirty per cent fewer offers. The outcome follows: 44.6% are employed after graduation, against 62.1% of domestic graduates.
Put the two effort figures together and the return on an evening of applying is roughly a third of what a domestic classmate gets for the same evening. That is not a motivation problem and it is not a skills problem. It is a filter that runs before either is assessed, and the person on the other side of it is doing more work for less result while being told, usually, to try harder.
This is the part of the story that generalises beyond the United States. Every market with a sponsorship regime has some version of this ratio, and it is almost never published, because it requires linking applications to outcomes rather than counting either alone.
| Measure | International | Domestic |
|---|---|---|
| Employed after graduation | 44.6% | 62.1% |
| Applications sent | Twice as many | Baseline |
| Offers received | 30% fewer | Baseline |
What this changes about how you spend an evening
If sponsorship appears on roughly one posting in thirty-eight, then for a candidate who needs it, the scarce resource is not effort and it is not resilience. It is knowing which postings those are before the evening is spent, rather than after.
The cost of the current arrangement is not that applications fail. It is that they fail late, after the CV was tailored and the cover letter was written, at the point in the form where the sponsorship question appears. Every one of those is an evening that a filter could have returned in a second.
What we can and cannot do about it
This desk belongs to a company that builds a job app, so the honest thing is to be exact about which half of that problem we solve.
We score every role on six published dimensions and show all six before you apply, so a role that is weak for you is visible without spending the evening to find out. That helps with allocating effort in general.
We cannot tell you whether a role offers visa sponsorship. We hold no sponsorship field, most job sources do not carry one in a structured form, and inferring it from free text would produce exactly the confident wrong answer that costs someone an application. Until we can read it reliably from the source rather than guess it, we will not show it, and we would rather say so than let a gap sit quietly behind a filter that looks complete.
The six dimensions this desk's company scores every role on, and the audit that tests them, are at babzituna.com/bias. Neither tells you whether a role offers sponsorship, and this edition says why we will not guess.