A writer or editor in AI training mostly judges drafts that a model wrote, and sometimes writes the reference answer it is judged against. The figures below come from the open postings on micro1, Ethos, AfterQuery, SME Careers and Mercor, read 13 October 2026.
Judging drafts is the main job; writing is the exception
Reviewing or refining AI-generated writing appears in 13 of the 16 full descriptions, across 16 open roles. The tasks fall into four kinds.
- Rating AI writing. micro1’s Journalist/Writer, Lawyer/Legal Writer and Educator/Assessor Writer postings ask you to compare several AI answers to a writing prompt. You score them and write a short rationale. They want you to spot prose that reads well but is wrong or unsupported.
- Writing the reference. micro1’s military and federal postings ask for “golden responses”, the model answers that other output is graded against, plus grading rubrics. The medical writer posting asks you to author evaluation tasks from clinical study reports and check AI drafts against ICH E3 structure. AfterQuery’s Content Writer and Evaluator posting asks for research tasks in areas such as sports analytics and patent research, not creative copy.
- Editing. The Senior Book Editor posting asks for manuscript feedback, line editing and copyediting, and for reviewing AI-generated editorial work.
- Business writing. Ethos has 5 postings for pitch decks, proposals and speeches. They ask you to create, evaluate and refine AI-generated slide decks. SME Careers lists a copywriting expert who rates and rewrites marketing copy.
Mercor’s news role fits the first kind. It asks reporters to compare AI-written news stories and scripts on reporting rigour and editorial soundness.
What each kind states it pays
| Kind of work | Roles | Advertised band, an hour |
|---|---|---|
| Rating AI writing, micro1 | 3 | $90–140 |
| Evaluating AI news, Mercor | 1 | $70–110 |
| Presentations and proposals, Ethos | 5 | $70–80 |
| Medical, military and federal writing, micro1 | 3 | $40–80 |
| Book editing, micro1 | 1 | $80–120 |
| Copywriting, SME Careers | 1 | up to $55 |
| Evaluation-scenario writing, AfterQuery | 1 | $15 |
The median ceiling across the 16 roles is $80 an hour. The spread is wide for similar work. AfterQuery’s evaluation posting states one flat rate, while micro1’s rating postings state a much higher band for comparable writing and evaluating work. The postings do not explain the difference.
Mercor’s news posting describes a different pay path. It pays on approval of a first task, then hourly if the first task is finished on time. It runs about four to six weeks and asks for 20 or more hours a week. SME’s copywriting posting states a ceiling only and says there is no immediate project. Qualified experts are contacted when one arises.
Proof of craft comes in five forms
- Years in the job. Postings name a number of years in 7 cases, usually four or five. Mercor and Ethos both ask for four.
- A portfolio. Ethos’s speechwriter posting asks for keynotes delivered at significant venues. The pitch deck posting asks for rounds closed from your materials.
- Published work. The book editor posting prefers experience editing published books for recognised houses.
- A degree from named countries. The Journalist/Writer posting prefers a university degree completed in the US, Canada, UK, Ireland, Australia or New Zealand. If yours is from elsewhere, check before you apply.
- Niche expertise. The medical writer posting prefers five years of regulatory medical writing and full clinical study reports. The military postings prefer Department of Defense documentation experience and a US base.
Who owns what you write
None of the 16 full descriptions mentions ownership, copyright, intellectual property, credit or royalties. Mercor’s data-policy page, read 13 October 2026, does not address who owns what contractors produce either. The terms sit in the contract you sign after selection. This site has not read any platform’s contract.
Your output is your craft, so ask in writing before you sign:
- Who owns the rewrites, reference answers and rubrics you produce, and may you reuse them?
- Will your name or work ever be credited, and may you list the project in a portfolio?
- Will anything you supply be used to train a model that competes with your own clients?
- Does the contract bar you from similar work for others?
Do not paste a client’s or employer’s unpublished material into any task. If you are on staff at a newsroom, publisher or agency, read your outside-work policy and any exclusivity clause in your freelance contracts. The federal reporting analyst posting asks for a record of confidential and compliant report handling.
Country limits
By their location field or their own wording, 5 of the 16 roles are limited to the US, including the military and federal postings. SME’s copywriting posting lists a spread of countries, including the US, Germany and several in South and Southeast Asia.
Which to try first
Start with a role whose entry bar you already clear. Federal Staff Reporting Analyst states the most openings in this group, at 50, and it is limited to the US. Without a niche background, the three micro1 rating postings are the broadest. They share a similar scope and differ in the professional background they prefer. Read what AI training work pays for how the bands compare. Read Ethos’s platform page before you apply, because one Ethos profile covers all of its roles.
What the postings do not say
- Who owns the work, and whether it can be credited.
- How many hours a week micro1’s roles offer, or how long they last.
- Which style guide or rubric applies before you start.
- Whether a talent-pool listing, such as SME’s, ever turns into a project.
- Whether using AI tools to draft your own rationales is allowed. Read the project rules first.
This site is not affiliated with any platform named here, and passing a screening is not a promise of work.
Open writing and editing roles right now
The 6 newest of 15 open roles this site lists today, with the pay each posting states.