Design and stress-test simulated content marketing workplaces — briefs, CMS states, dashboards and planted errors — that are used to evaluate AI agents on real editorial work. You write the reference answers and grading criteria, then judge agent attempts against your own standard.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

The work

Mercor is building high-fidelity replicas of a marketing org's tool surface — CMS, GA4, Semrush or similar, CRM, project management, social, ads, support — populated with realistic docs, dashboards, personas and deliberately planted problems. Your job is the Content Marketing slice of that: research and discovery, editorial prioritization, content production, and content QA and audit. A typical unit of work starts with a task you have actually done (a quarterly refresh decision, a cannibalization audit, a brief that conflicts with the analytics), then requires you to specify the environment it needs — which documents exist, what the GA4 view shows, what the CMS entry states say, which persona sends the Slack message that contradicts the brief.

The hard part is the planted issue. Every task must test something: a decision a strong practitioner reaches and a weaker one misses. You write the reference answer, then the criteria that separate a correct response from one that is fluent, confident and wrong — the plausible-but-wrong answer is the one the rubric has to catch. Later you review AI agent attempts and grade them against that standard, and where the agent surprises you, you say whether the rubric was wrong or the agent was.

What the screen looks for

Expect a short multiple-choice knowledge screener on content marketing specifics before human review, then an AI-led interview. It probes for hands-on ownership rather than management-by-summary: which CMS you were actually in, what your editorial calendar looked like, what you killed and why. Prior work on AI training environments, RL environments or simulated case studies is strongly preferred, as is Rubric Academy or Rubric Bootcamp certification. Written reasoning carries real weight — most of the value here is in explaining why a decision is right in a way another expert can audit.

Logistics

Fully remote and largely asynchronous, with work claimed in batches
Observed band $60–100/hr, stated as reported by contributors rather than guaranteed
Hours are flexible; most contributors treat this as part-time alongside practitioner work
Written deliverables dominate — briefs, environment specs, rubrics, grading notes

Marketing Expert — Content Marketing (AI Agent Environments)

The work

What the screen looks for

Logistics

Interview prep — 10 questions