What the work involves

You receive AI-generated artifacts — a press release, a media plan, a crisis comms memo, an editorial style audit, a stakeholder deck, a spreadsheet tracking coverage or audience metrics — and assess them the way you would assess a junior staffer's draft before it went to a client or an editor. That means checking factual claims and attribution, testing whether the argument holds, and flagging the things that quietly signal amateur work: invented sources, misused embargo or on-the-record conventions, headline formats no outlet actually uses, slides that bury the news in paragraph four.

Feedback is structured, not conversational. You score against a rubric, then write out what is wrong, why it matters in practice, and what a correct version would look like. Slide-heavy tasks are common, so real fluency in Google Slides and PowerPoint matters — you will often be judging layout hierarchy, chart honesty, and deck logic alongside prose quality.

What the platform screens for

  • Verifiable seniority. Five-plus years in newsrooms, agencies, in-house comms, or publishing, with specifics you can name: beats covered, outlets, campaign types, who you reported to.
  • Depth under follow-up. Screens push past your résumé summary into judgment calls — how you handled a correction, sourcing standards you enforced, what you cut from a deck and why.
  • Rubric discipline. Whether you can separate "I would have written it differently" from "this is factually or professionally wrong," and hold a consistent standard across many samples.
  • Written clarity. Your feedback is the deliverable; it is read as a work sample.

Logistics

Fully remote and asynchronous. Work is claimed from a task queue with deadlines rather than fixed shifts, so hours are flexible, but volume is project-driven and can be uneven week to week — treat it as supplemental rather than a guaranteed load. Most contributors work in blocks of a few hours; sustained batches tend to produce more consistent calibration. Pay bands reflect observed rates on the platform and vary by specialization, task complexity, and calibration performance.