Original research · last measured 2026-10-01
Who can actually apply to remote AI work
Remote AI-training and expert-data platforms describe themselves as global. Measured against 654 live listings that publish country eligibility, they are not — and the countries that lose out are not the ones you would guess. An applicant in India reaches 86% of this work; an applicant in Australia reaches 35%.
What the data says
There are three tiers, not two. The United States sits alone at the top: an applicant there can apply to 617 of the 654 measured jobs (97%). A broad middle band — the United Kingdom, Germany, Canada, Mexico, India, Brazil, Indonesia, the Philippines, Nigeria, Pakistan and South Africa — all land within a few points of each other, around 86%.
Then there is a floor, and it is occupied by countries most people would expect near the top. Australia reaches 35% of this work — roughly 330 fewer jobs than India, and the same band as Ireland, New Zealand, Sweden, Poland and South Korea. Being in a wealthy English-speaking country is not the advantage it looks like here; being in a country these platforms already recruit from is.
The cause is one list. The largest single restriction in the data covers 323 jobs — one platform publishes a 58-country eligibility list and applies it to every job it lists, and Australia, Ireland, New Zealand, Sweden, Norway, Poland, South Korea and China all fall outside it. That single list decides more of this catalogue than every per-job restriction combined.
The per-job restrictions are a smaller and more predictable story: of the 434 restricted jobs, 81 are open only to applicants in the United States, which is what puts the United States on its own at the top.
Jobs open by applicant location
Out of 654 jobs that publish eligibility. Countries shown are the largest remote-work populations plus common reference points; the full list runs to 102 codes and all of them except the United States fall within the range above.
| Applicant location | Jobs open | Share |
|---|---|---|
| United States | 617 | 97% |
| United Kingdom | 554 | 87% |
| Germany | 548 | 86% |
| Canada | 550 | 86% |
| Mexico | 546 | 85% |
| India | 550 | 86% |
| Brazil | 544 | 85% |
| Indonesia | 543 | 85% |
| Philippines | 543 | 85% |
| Nigeria | 543 | 85% |
| Pakistan | 543 | 85% |
| South Africa | 544 | 85% |
| Australia | 223 | 35% |
| Ireland | 224 | 35% |
| New Zealand | 222 | 35% |
| Sweden | 223 | 35% |
| Poland | 223 | 35% |
| South Korea | 221 | 35% |
How the restrictions are written
Restrictions are allow-lists, not blocks. A job names the countries it accepts and everywhere else falls outside it, so no country is ever listed as barred. 12 of the 22 distinct restriction patterns name a single country, covering 99 jobs between them.
| Restricted to | Jobs |
|---|---|
| 58 countries | 323 |
| United States | 81 |
| India | 6 |
| United Kingdom | 3 |
| Canada, United States | 2 |
| 5 countries | 2 |
| 40 countries | 2 |
| South Africa | 1 |
| Belgium | 1 |
| Switzerland | 1 |
What this does not measure
51 of the 705 jobs in this directory come from platforms that publish no eligibility field at all. They are excluded from every figure above rather than counted as open, because an absent field is not a statement that anyone may apply — it is the absence of a statement. Where a platform says nothing, the job page here says nothing too.
Published eligibility is also not the same as practical eligibility. A job open to your country can still require work authorisation, a tax identity the platform can pay against, or overlap with a client's working hours. This measures what the platforms publish, which is the part that can be measured.
Figures recomputed on every deploy from the live catalogue of 705 jobs (690 remote, 15 hybrid or on-site). If something here contradicts a platform's own page, the platform is right and this should be corrected.
Common questions
Can I do remote AI training work from India?
Yes, and India is among the better-served countries for it. Of the 654 jobs in this directory whose platform publishes country eligibility, 550 accept applicants in India — 86%. That is more than double the reach available from Australia, Ireland or New Zealand, which fall outside the eligibility list one major platform applies to all of its jobs.
Can I do remote AI training work from Australia?
Less than you would expect. Only 223 of the 654 measured jobs accept applicants in Australia — 35%, against 86% for India and 97% for the United States. Australia sits outside the 58-country list one platform applies to every job it publishes, and that single list covers 323 of the jobs here. Ireland, New Zealand, Sweden, Norway, Poland and South Korea are in the same position.
Are remote AI training jobs open worldwide?
No. 434 of 654 measured jobs restrict which countries may apply, and only 220 are published with no restriction at all. The largest single restriction is one platform's 58-country eligibility list, which it applies to all 323 of its jobs.
Which countries have the least access to remote AI training work?
Not the ones usually assumed. No country is named as barred — restrictions here are written as allow-lists, so a country is simply outside one. The countries with the least reach in this data are wealthy English-speaking and Northern European ones such as Australia, Ireland, New Zealand and Sweden, at roughly 35%, because they fall outside the largest platform-wide eligibility list. Countries these platforms actively recruit from — India, Nigeria, Pakistan, the Philippines, Brazil — reach around 86%.
Do I need to live in the United States for AI training work?
No. The United States leads at 97%, but the broad middle band sits around 86%, so the American advantage is roughly 67 extra jobs rather than a different category of access. The far larger divide is between countries inside and outside the platform-wide eligibility list.
How is this measured?
Every listing is refreshed daily from the platform's own public posting, and the country fields are read from that same payload rather than estimated. Both per-listing restrictions and platform-wide eligibility lists are counted, since both decide who may apply. Only the 654 jobs covered by a published eligibility rule are included; the other 51 come from platforms that publish none, and are excluded rather than assumed open — an absent field is not a clearance.
See all research, browse the 705 open jobs or read how this directory works.