What the work actually is

This is an open application to a talent network, not a posting for a defined job. Once you're verified, project invitations arrive on a rolling basis when an AI lab commissions marketing work. The tasks themselves tend to look like: writing realistic marketing prompts a model should be able to handle (a segmentation brief, a channel budget reallocation, a positioning statement), producing a reference-quality answer, then grading model outputs against it. Other engagements are pure evaluation — you read two model responses to a marketing scenario and justify, in writing, which is better and why the weaker one fails.

The judgment being purchased is yours as a practitioner. Models are fluent at marketing prose; they are much weaker at knowing when a CAC figure is implausible, when an attribution claim doesn't survive contact with how the data was collected, or when a brand recommendation ignores the category's actual competitive dynamics. Rationale quality matters more than verdict — reviewers read your written justification more closely than your score.

What the screen looks for

  • Verifiable specificity. Named channels, real platforms, budgets you actually managed, metrics you actually owned. Vague ownership claims get probed hard.
  • Depth on follow-up. Expect the interviewer to pick one claim from your resume and push three or four levels down into it.
  • Written clarity. Nearly all deliverables are written; the screen weighs how precisely you explain reasoning.
  • Independence. Little hand-holding, guidelines that change between projects, and a tolerance for ambiguity in rubrics.

Breadth across digital campaign management, market research and analytics, and content or brand strategy improves match rate, since project needs shift.

Logistics

Fully remote and largely asynchronous. Typical commitments are stated as 15–30 hours per week, though actual volume depends entirely on whether a project matching your background is live — expect gaps between engagements. The $60–80/hr band reflects observed rates in this network and is not guaranteed; individual projects set their own rates. Apply once, get verified, then wait for invitations rather than reapplying.