Two years ago, "AI training jobs" meant Mechanical-Turk-style microtasks paying a few dollars an hour. In 2026, the same label covers $40/hr audio recording gigs, $125/hr PhD-level reasoning evaluation, and $200+/hr domain-expert work for the major frontier labs. The market has bifurcated, and which platform you choose matters more than how many hours you put in.
Below is what each of the platforms aggregated on this site actually does — the kind of work you'll be doing, the pay band you can realistically expect, and who they're for. All listings on our home page route through these platforms directly; we earn a referral commission on some signups, which keeps the listings free.
The frontier-lab tier ($60–$200+/hr)
These platforms recruit subject-matter experts for the labs training Sora, Claude, Grok, Gemini, and similar systems. Selection is competitive and roles are typically time-limited (weeks to a few months), but pay is well above market.
- Mercor — the most prolific gig listings in the space (~50 active roles at any time) covering equity research analysts, lawyers, doctors, software engineers, and STEM specialists. Hourly contracts, $40–$200/hr, with a referral program that pays 20% of the referee's earnings up to 4× their hourly rate. Read our full Mercor review.
- micro1 — almost 400 listings spanning generalist evaluators, code reviewers, and niche professional specialists (insurance, manufacturing, nursing, pixel art). Hourly $20–$70 for most roles, $50–$150 for engineering. The referral program pays a flat $3,000 per hired referral who works 10+ hours.
- Handshake AI Fellowship — 200+ part-time opportunities aimed at master's, PhD, and postdoc candidates. Up to $125/hr. Fully remote, U.S. work-authorization required. Fields range from 2D design and aerospace engineering to air traffic control and archival research.
- Welo Data (Welocalize) — ~480 active hubs for linguists, audio annotators, and cultural specialists across dozens of languages. Most postings list specific language requirements; pay scales with locale and specialization.
The voice & audio tier ($10–$50/hr, easier to get into)
Lower bar to entry, plenty of work, and the application process is usually a short voice sample rather than a multi-stage interview.
- Babel Audio — recorded conversations and voice prompts in dozens of languages. The English roles pay around $17–$50/hr; some non-English roles (German voice acting, for instance) post rates as high as $150–$225/hr. Apply once and they pull you into projects as they open.
- Pila8 — solo and conversational audio recordings. $10–$50/hr depending on the project, payable in 14+ languages. Quick onboarding, no AI interview, payment via Dots (PayPal/Venmo/CashApp/local bank).
- CrowdGen by Appen — Appen's consumer-facing crowd platform. ~50 active listings on their public board, mostly accented audio contributors and search relevance evaluators. Predictable but not generous pay (often $15–$25/hr).
The microtask & data-collection tier
- OneForma (Centific) — task-based work spanning dermatology projects, French Canadian LLM evaluation, and various data collections. Variable pay, variable availability, but consistently several dozen open projects.
- Remote Talent Cloud — U.S.-only AI Training Specialist roles around $20/hr. Fully remote, steady hours, basic data annotation work.
- xAI (the makers of Grok) — alongside corporate engineering hires, xAI posts a steady stream of "tutor" roles (3D Tutor, Video Games Tutor, etc.) on their public Greenhouse board. These tend to be contract gigs at $40–$100/hr.
What we don't list (yet)
Several major platforms — Outlier, Alignerr, AfterQuery, Atlas Capture, Turing, DataAnnotation.tech, Toloka, RemoteExperts, RWS TrainAI, Pareto AI, SME Careers (SuperAnnotate), Mindrift, Prolific, TELUS International AI, Uber AI Solutions — keep their listings behind a login wall, hydrate them client-side, or sit behind Cloudflare bot protection. We're building headless-browser ingestion for those next. If you have a specific platform you want bumped to the top of that queue, let us know.
How to choose
The honest answer: most active contributors work across three or four platforms simultaneously. Frontier-lab gigs (Mercor, micro1, Handshake) are higher-paying but project-driven and bursty — work appears, then disappears for weeks. Audio platforms (Babel, Pila8) are steadier but capped at lower rates. The combination keeps your hours full.
If you're starting out, we'd apply to Mercor and Babel Audio first — Mercor for the upside, Babel for the immediate-week work — and layer in micro1 and Handshake AI once you have a track record. The listings page is updated daily and filterable by platform, pay range, commitment, and country eligibility.