Guide

How to get accepted on AI training platforms

What Mercor, micro1, Outlier, DataAnnotation, Handshake AI and Turing ask for, which CV rules are official and which are only rules of thumb, and how to prepare for AI interviews.

AI-training platforms decide fast, often with an automated screen and an AI interview, and many give you only one attempt. This guide separates what the platforms officially say from rules of thumb that circulate online. Each tip says which it is. Rules change often, so each platform’s page on this site shows the current, dated facts.

Six rules that apply everywhere

  1. Never invent anything. No made-up numbers, tools, titles or credentials. Platforms verify, and interviews expose inflated CVs.
  2. Keep confidential work confidential. Don’t name internal project codenames or clients from earlier AI-training work. Describe the work instead: “STEM content quality evaluation”, “prompt adjudication”.
  3. Share as little personal data as you can. City and country, not your street address. No date of birth or ID numbers unless a listing asks. AI-hiring platforms hold sensitive data, and a reported 2026 data breach at one of them was followed by lawsuits that are still pending.
  4. Write in English and list your languages with levels: native, fluent, professional working, conversational. Language pairs are their own job category.
  5. Lead with the field you can prove. Expert tracks pay several times more than generalist tracks. In October 2026, micro1’s median listed pay ranged from about $28/h for languages to about $150/h for healthcare (aitrainer.work dataset, CC BY 4.0).
  6. Use a text-based PDF: no scans, no text inside images, no complex tables. Parsers read text, and Mercor requires OCR-friendly PDFs (Mercor docs).

Mercor: what’s official and what’s a rule of thumb

Official (Mercor’s talent docs, checked October 2026):

  • Résumé: PDF only, OCR-friendly text, 1–2 pages, under 2 MB (How to apply).
  • Instant offers come from your match score: experience, past interviews and assessments, skills and availability. Offers expire, and declining one doesn’t hurt future matching (Instant offers).
  • AI interview: only your latest attempt counts, and a retake resets that application’s progress (How to apply).
  • Passed assessments are reused for comparable roles (How to apply).
  • Hourly work needs the Insightful desktop app, so no Chromebook, iPad or phone (Mercor talent docs).
  • Real offers arrive inside the Mercor platform. Messages elsewhere from “Mercor recruiters” may be impersonation (Referral policy).

Rules of thumb (common advice, not confirmed by Mercor):

  • A summary of about 150–200 words: who you are → your current work with named tools → two or three results with numbers → the roles you want, time zone, availability and languages. Good writing for any reader.
  • A tagline like “[Area 1] & [Area 2] | [What sets you apart] | [X+ years]”.
  • Put your strongest standard skill terms first. Claims that the parser reads only the first five skills are unconfirmed.
  • An exact “Skills & Tools” heading. Harmless, but unconfirmed.

Platform by platform

micro1

  • An AI interviewer runs a 20–30 minute interview with camera and screen sharing. An integrity score uses gaze and tab focus, and one attempt counts (aitrainer.work, October 2026).
  • Prepare two or three stories from your field with numbers, plus one example of catching an error. Quiet room, camera at eye level, every other tab closed, screen sharing tested first.
  • About 20 countries are excluded, including China, Russia, Iran, Venezuela and Ukraine (same source).

Outlier (Scale AI)

  • Claim only the two or three domains you can prove. Specialist tracks pay more but test harder.
  • Onboarding and project qualifications are unpaid, and pay is shown per project before you start (The AI Rankings).
  • Keep project details confidential. Account bans are the most common complaint in community forums, so don’t risk one.

DataAnnotation

  • The Starter Assessment is unpaid, with one attempt and no feedback. Do it rested, follow the instructions exactly and explain your reasoning (The AI Rankings).
  • Reported eligible countries are the US, Canada, the UK, Ireland, Australia and New Zealand; the company doesn’t publish a list (aitrainer.work).

Handshake AI

  • Official eligibility: US-based students, graduates and PhD candidates with work authorization, plus ID verification and a background check (Handshake support).
  • Fill in your degree, field, coursework and research: projects are matched by field.

Turing

  • Timed coding or analytical and writing assessments. Most listings that state a policy require a video interview, and many need partial overlap with US Pacific time and about 20 hours a week (aitrainer.work).
  • Emphasise shipped systems, languages and frameworks, and measurable engineering results.

AfterQuery, Ethos and expert networks

  • AfterQuery: emphasise years of hands-on practice and the kinds of decisions you make; the work is documenting expert reasoning.
  • Ethos: a single AI voice interview builds your profile (The Next Web). Prepare a crisp story of the companies, roles, markets and products you know.
  • Expert networks: precision beats prestige: exact companies, products, regions and years. Never share confidential information about an employer (Arbolus).

AI interview checklist

  1. Test camera, microphone, screen sharing and browser before you start.
  2. Use a quiet room, a neutral background and a stable connection.
  3. Structure answers: the situation → what you did → the result, with a number → what you’d do differently.
  4. Don’t use AI assistants during a proctored interview. It breaks platform rules and integrity scoring.
  5. Know the retake rules first: on Mercor a retake resets your progress; on micro1 one attempt counts.

Pages mentioned in this guide

  • Step 1 · Paid tasks Turbulent status

    Mercor

    The largest expert marketplace for AI labs: hourly contracts for professionals, from generalists to doctors, lawyers and senior engineers.

    Realistic pay ≈ $16–150/h
    First money 7–60 days
    Cost to start $0
    Grade A: Official source Grade A means: Official source. The grade says how well-sourced this page is, not how good Mercor is. More in the glossary Official source Checked 4 days ago
  • Step 1 · Paid tasks Open status

    micro1

    AI-interview-gated talent pool ('Zara') placing experts on AI-lab projects; strong rates for STEM, legal and healthcare.

    Realistic pay ≈ $25–150/h
    First money 14–60 days
    Cost to start $0
    Grade B: Reliable source Grade B means: Reliable source. The grade says how well-sourced this page is, not how good micro1 is. More in the glossary Reliable source Checked 4 days ago
  • Step 1 · Paid tasks Turbulent status

    Outlier (Scale AI)

    Scale AI's contributor platform: rate, write and fix AI outputs, with specialist tracks up to ~$150/h advertised.

    Realistic pay ≈ $12–60/h
    First money 7–30 days
    Cost to start $0
    Grade B: Reliable source Grade B means: Reliable source. The grade says how well-sourced this page is, not how good Outlier is. More in the glossary Reliable source Checked 4 days ago
  • Step 1 · Paid tasks Turbulent status

    DataAnnotation (Surge AI)

    Pick-your-own-hours AI tasks with fast PayPal withdrawals; one unpaid entry assessment decides whether you get work.

    Realistic pay ≈ $14–45/h
    First money 3–30 days
    Cost to start $0
    Grade B: Reliable source Grade B means: Reliable source. The grade says how well-sourced this page is, not how good DataAnnotation is. More in the glossary Reliable source Checked 4 days ago
  • Step 1 · Paid tasks Turbulent status

    Handshake AI Fellowship

    AI training projects for US-based students, grads and PhDs via the Handshake careers network.

    Realistic pay ≈ $30–125/h
    First money 14–45 days
    Cost to start $0
    Grade B: Reliable source Grade B means: Reliable source. The grade says how well-sourced this page is, not how good Handshake AI Fellowship is. More in the glossary Reliable source Checked 4 days ago
  • Step 1 · Paid tasks Open status

    Turing

    AGI-infrastructure company hiring contractors (mostly engineers, increasingly other domains) for lab projects; often needs PST overlap.

    Realistic pay ≈ $25–90/h
    First money 14–28 days
    Cost to start $0
    Grade C: Self-reported Grade C means: Self-reported. The grade says how well-sourced this page is, not how good Turing is. More in the glossary Self-reported Checked 4 days ago
  • Step 1 · Paid tasks Open status

    AfterQuery

    Fast-growing YC startup paying working professionals to document how they reason through real cases.

    Realistic pay ≈ $50–150/h
    First money 14–45 days
    Cost to start $0
    Grade B: Reliable source Grade B means: Reliable source. The grade says how well-sourced this page is, not how good AfterQuery is. More in the glossary Reliable source Checked 4 days ago
  • Step 1 · Paid tasks Open status

    Ethos

    London-based AI expert network (a16z-backed): one voice interview builds your profile; matched to expert calls, surveys and AI-data work.

    Realistic pay ≈ $60–225/h
    First money 14–60 days
    Cost to start $0
    Grade B: Reliable source Grade B means: Reliable source. The grade says how well-sourced this page is, not how good Ethos is. More in the glossary Reliable source Checked 4 days ago
  • Step 2 · Your skill Open status

    Expert networks (GLG, AlphaSights, Guidepoint, Third Bridge, Arbolus …)

    Paid 30–60 min calls and surveys for investors and consultants who need your industry knowledge; senior niche expertise earns $300–1,000+/h.

    Realistic pay ≈ $100–400/h
    First money 7–120 days
    Cost to start $0
    Grade C: Self-reported Grade C means: Self-reported. The grade says how well-sourced this page is, not how good Expert networks is. More in the glossary Self-reported Checked 4 days ago