Guide · getting started
How to get AI training jobs
By Alex. First published 2026-09-01. Pay figures as of 2026-10-25, from live platform feeds.
AI training jobs are contract roles where you rate model answers, write reference responses, check code, or lend expertise in a field you already have. The companies hiring are real. The listings are scattered, they close fast, and most "salary" pages quote an undated span with no sample size. This site reads the platforms' own public job feeds. Right now that is 3,292 open roles, with typical top rates of $34–$110/hr (median $70/hr, from 545 distinct listed rates across 25 platforms).
I do this work myself, on top of a day job. The steps below are what I wish I had followed from day one, not a promise that you will be accepted on the first try.
1. Decide what you can actually do
You do not need an AI degree. You do need something you can explain: a profession, a language, a coding stack, or the patience to follow a rubric. Platforms test judgment and instruction-following more than they test machine-learning trivia.
- Model evaluation & rating currently shows a median top rate of $40/hr (typical $25–$60/hr, highest $200/hr, from 53 distinct rates, on a page listing 226 live roles). Anyone who can read carefully can try.
- Writing & content currently shows a median top rate of $55/hr (typical $40–$80/hr, highest $140/hr, from 24 distinct rates, on a page listing 46 live roles).
- Software engineering currently shows a median top rate of $90/hr (typical $65–$120/hr, highest $400/hr, from 138 distinct rates, on a page listing 471 live roles).
- Law & legal currently shows a median top rate of $125/hr (typical $100–$150/hr, highest $400/hr, from 48 distinct rates, on a page listing 99 live roles).
Rating & evaluation roles →Coding evaluation roles →Domain-expert roles →
2. Apply to several platforms the same week
Do not pick one brand and wait. Queues go quiet without warning, and a single rejection is normal. I applied many times before I was accepted on more than one platform. The live feeds on this site currently cover 29 platforms that publish public opportunities. Outlier's project listings include country restrictions; DataAnnotation's rows are ongoing recruitment tracks. Neither a listing nor acceptance guarantees available tasks.
- Open the live list. It shows the newest roles first, because fresh postings fill first.
- Apply to every role you can honestly explain. An imperfect match is still worth an application if you meet the essential requirements.
- Keep a master CV and a 60-second spoken summary of your experience. Almost every AI interview opens with that question.
- Turn on alerts so you see a role the day it appears, not the day it closes.
Live jobs, newest first →Email alerts →Every platform we track →
3. Treat the interview as a reusable asset
The first AI interview is the slow one. Later applications on the same platform often reuse it. I have written up what that actually felt like on Mercor, micro1 and Ethos, including the exercises, screen sharing and the parts that wasted my time.
What happens after you apply (first-hand) →
4. Do not live on one platform
These are contracts, not jobs with a manager who owes you hours. A platform can be full on Tuesday and empty on Friday. People who stick with this work usually keep two or three platforms active so a dry week on one does not zero the week.
Listed rates are the top of a band, not a guaranteed wage, and a role must show at least $10/hr to appear here. Empty queues, unpaid training and failed quality checks all pull take-home down. Read the listing's own wording on the job page before you invest an evening in an assessment.
Common questions
- Do I need AI experience to get an AI training job?
- Usually no. Platforms test whether you follow a rubric, write clearly and judge answers. A profession, a language or a coding background unlocks better-paid work, but general rating roles exist for people without any of those.
- How long does it take to get accepted?
- Anywhere from a day to more than a month, and many applications never get a yes. That ratio is normal. Applying early to fresh listings is the part you can control.
- How much do AI training jobs pay?
- On this directory right now, typical top rates of $34–$110/hr (median $70/hr, from 545 distinct listed rates across 25 platforms). That describes listed roles (listed at least $10/hr), not every gig on the internet. The pay index updates with the feeds.
- Which platform should I start with?
- Start with whichever live listing matches work you can explain, not with a brand name. If you want a named starting point, apply to two feed platforms and one profile-only platform in the same week so you are not waiting on a single queue.