Both, honestly. There are real platforms paying real people to help train, evaluate and apply AI models, and the hourly rates on public sources are not fantasy. But most of the loud content about "AI income" online is selling you a course, a template pack, or an affiliate link. The four platforms below are free to join and pay you directly.
What's true: the demand for people who can work with AI is growing fast, and beginners can enter without a degree or a portfolio. What's also true: the first dollar is the hardest, work is inconsistent, and headline rates usually belong to people with existing professional expertise. If you're expecting a steady salary in week one, you'll be disappointed and quit. Treat this as skill-building with variable income attached, not as a job replacement.
Hourly rates and time-to-first-payment as reported by the sources listed at the end of this article.
DataAnnotation — you review and correct AI model outputs. No degree needed. Base pay commonly reported at $20/hr, with specialist and coding tasks reported higher (source). You must pass an unpaid assessment first, and acceptance is not guaranteed.
Mercor — a marketplace that connects domain experts to AI labs. Rates reported from ~$25/hr up to $200/hr for scarce professional expertise like medicine, law and finance (source). Top rates go to specialists; if you have no professional domain, this is a month-3 target.
Upwork — general freelance marketplace. Published rate guidance: entry-level $10–$25/hr, intermediate $25–$75/hr, specialised AI/development/consulting $75–$150+/hr (source). Free to join; the hard part is winning your first contract.
Fiverr — you list a fixed offer, buyers find you. Pay is highly variable; sellers only see meaningful volume after several completed orders build their ranking (source).
No sign-up, nothing to install, and it takes about a minute. This is the same thing the course has you do on day one.
You'll end up with: A proposal you can send to charge more.
Runs on bloom0's own AI. Nothing you type is stored or shared.
All four are free to join. What they cost is your time and your patience.
The honest picture: expect 1–4 weeks before any money arrives on Upwork or Fiverr, days-to-weeks on DataAnnotation, weeks on Mercor. Plan accordingly.
Pick DataAnnotation if you want the simplest path, don't have a professional specialism yet, and are comfortable with an assessment that might reject you. Good English and careful reading matter more than technical skill.
Pick Mercor if you already have real professional experience — nurse, lawyer, accountant, engineer, translator — and you want that experience to be the thing you're paid for. If you don't yet, come back to it later.
Pick Upwork if you're willing to learn one narrow AI-adjacent skill (writing with AI, building simple automations, cleaning data, setting up chatbots) and pitch it. The narrower the offer, the faster the first contract.
Pick Fiverr if you'd rather package a repeatable service and let buyers come to you. Works well for small, defined outputs: a logo variant pack, a translated document, a résumé rewrite, a set of product descriptions.
Many people start with DataAnnotation for cashflow and build an Upwork or Fiverr presence in parallel.
You fail the DataAnnotation assessment. This is the most common failure point. Read every instruction twice, don't rush, and treat the assessment as the job — because it is.
You bid on Upwork jobs you can't clearly do. Generic proposals get ignored. A specific, short proposal that names the client's actual problem beats a long one every time.
Your Fiverr gig gets no views. Almost always because the gig is too broad ("I will help with AI") instead of specific ("I will write 10 SEO product descriptions from your spec sheet").
You apply to Mercor with a thin profile. Mercor's value is verified expertise. If your CV doesn't clearly show a domain, the match won't happen.
You quit in week two. The biggest hidden reason people "can't make money with AI" is that they stop after two silent weeks. The people who earn are the ones still there in week five, having adjusted their offer twice.
You'll be more competitive on every platform above if you can actually build something small with AI — a working chatbot, a simple automation, a clean prompt system for a specific task. That's what turns you from "another beginner" into "the person who solved my problem".
That's what bloom0 exists for. It's a free bilingual course that takes complete beginners from zero to building real things with AI. It won't hand you clients, but it gives you something concrete to point to when a Fiverr buyer or Upwork client asks what you can actually do. A profile that says "I built X" outperforms one that says "I know AI" every time. Free is free — no card, no upsell. Use it in parallel with the plan below.
Total time: about 315 minutes.
Three questions. No score is stored, nothing is sent anywhere.
1. You have no professional specialism yet and want money soonest. Which is the sensible first move?
Mercor's top rates go to verified specialists, so it's a poor first move without a domain. Broad Upwork offers get lost against thousands of others. DataAnnotation gives you the shortest realistic path to a first payment, and a narrow Fiverr gig runs in the background while you wait.
2. You've sent 15 Upwork proposals in two weeks and heard nothing. What's the most likely fix?
Volume without specificity is the classic new-freelancer mistake. Clients skim. A short proposal that names their problem and shows you understood the brief beats a generic one at any price. Cutting your rate alone usually attracts worse clients, not better ones.
3. Which statement about earnings on these platforms is honest?
The sources cite hourly ranges, not guaranteed monthly income. Work availability on DataAnnotation is inconsistent, Mercor's top rates require expertise, and Upwork/Fiverr income depends on winning your first jobs. Treating these as variable income rather than a salary is the mindset that keeps people going long enough to actually earn.
Free forever · no account needed · nothing you make is shared