What the work involves

You receive a set of prompts drawn from real questions of UK corporate and commercial law — share purchase mechanics, warranties and indemnities, Companies Act 2006 requirements, Takeover Code points, drag/tag provisions, filings and consents. You run each prompt through two separate LLM platforms, then read both outputs closely and score them against a standardized rubric the client provides. Every evaluation is accompanied by short written feedback: what the model got wrong, whether the error is a citation problem or a substantive misstatement of law, and what a client would actually have been misled about. Volume matters less than the quality of the reasoning trail you leave behind.

The useful skill here is not spotting an obviously bad answer — it's calibrating between two answers that are both plausible and mostly correct. Reviewers who add value can say why one drafting suggestion is market-standard and the other is not, notice when a model quietly applies US or EU concepts to an English-law question, and distinguish an answer that is incomplete from one that is wrong.

What the screening looks for

  • Verifiable practice history — where you qualified, which firm, what kind of deals, and how recently. Law-firm transactional experience is the gate; recent in-house moves are acceptable only if the firm years came first.
  • Depth under follow-up — the screen will push past your first answer on a corporate law point to see whether the detail holds.
  • Rubric discipline — whether you can score to someone else's criteria rather than substituting your own instincts, and whether you can justify a score in two sentences.
  • Written precision — feedback is the deliverable, so vague comments score poorly.

Logistics

Fully remote and asynchronous; you choose when to work within the deadlines set for each batch. The initial engagement is roughly 10 hours, and reviewers whose feedback holds up are typically offered continuing batches. All prompts, rubrics and platform outputs are covered by NDA. Pay in the $150–160/hr range has been observed on comparable Mercor legal benchmarking projects; it is set by the client and not guaranteed.