AI Work
Read what a model wrote. Decide if it's right.
AI labs pay for careful human judgment on model output. You compare two answers and explain which is wrong and why. You score summaries against written rubrics. You catch the confident sentence with a fabricated citation in it. The skill being bought is skepticism, not technical background.
$85/hr general evaluation · $105/hr rubric and comparison · paid 2× a week
What the work is
A task from this queue.
Compare two answers about a visa rule
Both answers cite sources. Answer B is longer, better formatted, and cites a government page. One of its claims is not on that page. Pick the better answer, then name the specific claim that is wrong. Estimated time: about 9 minutes.
The numbers.
| General evaluation | $85/hr |
| Rubric & comparison | $105/hr |
| Specialist domains | $135/hr |
| Assessment | 6 questions · 20 min |
| Decision | within 72 hours |
| Minimum hours | none |
Worth knowing
Before you apply.
No degree required. The assessment measures one thing: do you verify claims, or do you trust formatting. That is the whole credential.
Written English at a professional level is required for English-language projects. Other languages open specific queues, so list every language you work in when you apply.
AI Work is selective. Strong verification answers with weaker model-judgment answers get you an offer for Web Work instead of a decline, and your decision email explains the routing.
Specialist queues in law, medicine, finance, and advanced STEM pay $135/hr. They require demonstrated domain knowledge in a follow-up sample, not a certificate.
Apply
6 questions, about 20 minutes.
No resume at any point. Decision by email within 72 hours.