Every answer ChatGPT, Claude or Gemini gives you was shaped by a human being who told the model what a good answer looks like. Those people are paid by the hour, and the doors to that work (Outlier and DataAnnotation) are open to anyone who can write clearly and think carefully. This guide shows you exactly how to apply, how to pass the assessment, how to avoid the scams around this niche, and, because many readers live in countries where PayPal does not work, how to plan your payment route before you invest any time.
- What AI training work actually is
- The honest reality check (read this first)
- Outlier vs DataAnnotation at a glance
- Step 0: Prepare before you apply
- Step-by-step: getting started on Outlier
- Step-by-step: getting started on DataAnnotation
- How to pass the assessments
- How you get paid (and the PayPal problem)
- Realistic earnings math
- Scam red flags in AI training jobs
- Free video walkthrough
- FAQ
What AI training work actually is
Large language models learn from huge amounts of text, but raw text does not teach a model what a helpful, accurate, safe answer looks like. That comes from human feedback. Platforms hire contractors to do tasks such as:
- Ranking responses: the model writes several answers, and you order them from best to worst and explain why.
- Writing hard prompts: you invent questions that trip the model up, then write the correct answer.
- Fact-checking: you catch wrong claims, made-up sources and logic errors.
- Writing or editing answers: you produce model-quality responses in your own words.
- Reviewing code or specialist content: programmers, mathematicians, lawyers, doctors and scientists check technical answers.
This is why the work pays better than basic microtasks: it needs judgment, not just clicking.
The honest reality check (read this first)
- Work is inconsistent. Both platforms depend on client demand. Contributors regularly report empty task queues and projects that end without notice.
- Assessments are unpaid. Expect to spend one to several unpaid hours on tests, and some contributors wait weeks or longer before seeing a project.
- Acceptance is selective. DataAnnotation’s own blog has been reported to put its acceptance rate at roughly 2–3%, and its starter assessment can be taken only once.
- Quality scores matter. Low-quality or rushed work can cost you access to projects.
- Country rules apply. Availability, pay and payout methods can differ by country, so check them first (see the payment section).
None of that means the platforms are scams. Both pay real money. It means you should plan for them as one income source among several, and apply once, carefully.
Outlier vs DataAnnotation at a glance
| Feature | Outlier | DataAnnotation |
|---|---|---|
| Run by | Scale AI | DataAnnotation (privacy notice names no separate legal entity) |
| Cost to join | Free | Free |
| How you start | Create account, pick a field, verify identity, take project onboarding/assessment | Pick a track, take one Starter Assessment (about 1–2 hours), wait for email |
| Retake allowed? | Varies by project | No, the Starter Assessment is one attempt |
| Pay model | Per project; rate shown before you accept a task | Mostly hourly with a built-in timer, some flat-rate tasks |
| Published/advertised pay | Varies widely by project, country and language; contributor reports commonly fall around $12–30/hr for standard work, more for verified specialists | General work from about $25/hr; coding from about $40/hr; STEM and professional tracks from about $40/hr (platform FAQ) |
| Payout methods | PayPal, Airtm or bank transfer (ACH), depending on region | PayPal |
| Pay schedule | Weekly (Tuesdays) | On request, delivered within a few days |
| Country access | Many countries, but no official public list; some projects are country-restricted | Historically limited to a short list of countries; check current rules at signup |
| Best for | Writers, language and subject experts, developers | Strong writers, coders, STEM and professional experts |
Figures come from each platform’s own pages and contributor reports; they change often, so confirm current numbers inside your dashboard.
Step 0: Prepare before you apply
Because DataAnnotation gives you a single attempt, preparation is the highest-return hour you will spend.
- A valid, unexpired government ID with your photo (passport or national ID card).
- A smartphone with a working front camera (identity checks use a selfie).
- A stable internet connection and no VPN on any device you will use.
- An up-to-date CV and LinkedIn profile that match each other.
- A quiet 2-hour block, a notes document, and a browser for fact-checking.
- Your payout plan decided in advance (see the payment section).
Also pick one track that fits you honestly: general writing and reasoning, a language track, coding, or a specialist field. Claiming to be good at everything usually loses to showing real depth in one area.
Step-by-step: getting started on Outlier
- Visit the official site only. Go to outlier.ai and open the list of opportunities. Type the address yourself rather than clicking links from messages or social media.
Screenshot: Outlier’s official home page (outlier.ai) - Create your account. Use an email you check daily. Sign-up is free; if anyone asks for a fee, stop.
- Choose your area of expertise. Be specific: coding, math, science, a language, writing, or a professional field. This is how you get matched to projects.
- Verify your identity. You will upload ID and may take a selfie. Do this on a personal device without a VPN. A mismatch between your stated country and your connection is a common reason for rejection.
- Take the project onboarding and assessment. General onboarding commonly takes 30–90 minutes, and each project can have its own guidelines and test. Read the instructions twice before you answer.
- Accept a project and read the guideline document. Most low-quality ratings come from people who skipped the guidelines. The hourly or task rate is shown before you start.
- Set up payments early. Connect PayPal, Airtm or a bank account in the payments section, and withdraw each week instead of letting a balance sit.
Outlier does not publish a definitive list of eligible countries, so the practical test is simple: create an account and see whether projects appear for your location. If none do, that is your answer, and you should not try to get around it with a VPN or false details.
Step-by-step: getting started on DataAnnotation
- Read the official FAQ first. Open dataannotation.tech/faqs. It explains tracks, pay, identity verification and the one-attempt rule.
Screenshot: DataAnnotation’s official website (dataannotation.tech) - Choose your track. You pick a Starter Assessment such as general, coding, multilingual or a specialist field. Coding requires real programming skills, and STEM and professional tracks require advanced degrees or credentials.
- Block out uninterrupted time. Most Starter Assessments take around one hour, and specialist ones one to two hours. Do not start it on a tired evening.
- Submit once, carefully. There are no retakes. Re-read every answer before submitting.
- Check your email. The platform says it emails you within a few days if you are approved. Many applicants hear nothing, which usually means they were not accepted.
- Complete identity verification when prompted. It uses a third-party service and needs a physical ID plus a selfie. It is free, and the company states it will never ask you for money.
- Take extra qualifications. After approval, additional qualifications on your dashboard open more projects. Take them seriously, since they unlock more work.
How to pass the assessments
1. Write like a careful editor, not a fast typist
Assessments reward thorough, accurate, well-reasoned answers. Give a clear reason for every judgment you make. One-line answers read as low effort.
2. Fact-check every claim
If the AI says something confident, verify it. Plausible-sounding errors are exactly what you are being tested to catch. Rating a wrong answer as correct can disqualify you.
3. Follow instructions to the letter
If the task says to justify in two sentences, do that. Format, length and tone instructions are part of the test.
4. Do not use AI to write your answers
Using AI to complete an AI-training assessment can get you rejected or banned. Your own reasoning is what is being measured, so use AI tools only if the platform’s rules explicitly allow it for that task.
5. Keep quality high on your first real tasks
Your early scores influence which projects you are invited to later. Speed comes with practice; accuracy has to come first.
How you get paid (and the PayPal problem)
This is the part most beginners discover too late. PayPal does not operate for residents of Pakistan, and DataAnnotation’s FAQ says it pays through PayPal. If you live in a country where PayPal is unavailable, check your payout route before you spend hours on an assessment.
- Do not use someone else’s identity, ID or PayPal account.
- Do not use a VPN to pretend you live in another country.
- Do not buy or rent a “verified account” from a stranger.
Payout options at a glance
| Platform | Methods | What to check |
|---|---|---|
| Outlier | PayPal, Airtm, bank transfer (ACH) | Open the payments section in your account and see which methods are offered for your country. Airtm is designed to receive USD and cash out to local currency, but cash-out options depend on your location, so confirm inside the Airtm app. |
| DataAnnotation | PayPal | If PayPal is not available where you live, this platform may not be practical for you. |
Outlier’s own payment announcement lists PayPal, Airtm and bank transfer as supported providers and says it has improved its earnings tab and support for payment issues (see Outlier’s post). Always confirm the current options in your dashboard, since terms change.
- Withdraw weekly, never let a balance accumulate.
- Keep a spreadsheet of hours, task IDs and pay so you can dispute a mistake.
- You are an independent contractor, so you are responsible for tracking and declaring your own income under your local tax rules.
Realistic earnings math
Advertised rates are not what lands in your account. Unpaid assessments, task-limit gaps and per-task pay that works out lower than expected all reduce your real hourly figure. Here is a simple planning model with illustrative numbers, not promises:
| Scenario | Hours per week | Nominal rate | Planning factor | Rough weekly result |
|---|---|---|---|---|
| Cautious | 5 | $15/hr | 60% | about $45 |
| Typical side income | 10 | $20/hr | 70% | about $140 |
| Skilled specialist | 10 | $40/hr | 80% | about $320 |
The planning factor accounts for unpaid onboarding, dry weeks and the gap between per-task pay and true hourly pay. Contributors who do best often run more than one platform and work on whichever has tasks that week. If you are exploring several, our guides on Prolific, CrowdGen by Appen and Clickworker cover other options, and our AI freelancing guide shows how to combine them with client work.
Scam red flags in AI training jobs
Because “AI training” is popular, scammers copy the names of real companies. Walk away if you see any of these:
- You are asked to pay for training, equipment, software, onboarding, an “account activation” or “payment release.”
- Someone sends you a check and asks you to buy equipment or forward money.
- A recruiter messages you on WhatsApp or Telegram and promises instant hiring.
- The role is vague (“AI tutor”, “digital trainer”) with no details about the actual work.
- Your first “earnings” appear in a strange app that asks you to deposit money to withdraw them (a task scam).
- The sender’s email domain does not match the company’s official domain.
The rule is simple: real platforms pay you; they never charge you to start. If you want a refresher on spotting fake earning apps, read our guide on how to spot fake money-making apps.
Free video walkthrough
If you learn better by watching, this free YouTube video explains how people earn from Outlier-style AI training work. It is embedded from the creator’s own channel, so views and credit go to them. Treat any income numbers in videos as examples, not guarantees.
Common mistakes that cost people their access
- Rushing the assessment because it feels like a quiz instead of a job interview.
- Skipping project guidelines and submitting work that ignores the rules.
- Using a VPN or a shared computer during verification.
- Relying on a single platform for rent or bills.
- Sharing task content publicly. Projects are usually under confidentiality terms, so do not post screenshots of live tasks.
- Ignoring deactivation emails or quality feedback instead of adjusting.
Frequently asked questions
Are Outlier and DataAnnotation legit?
Both are real businesses that pay contractors. DataAnnotation’s FAQ says it has paid over $20 million since 2020, and Outlier is operated by Scale AI. The common complaints are about inconsistent work and slow support, not about people never being paid.
Do I need coding skills?
No. General tracks focus on writing, reasoning and fact-checking. Coding tracks do require programming experience and usually pay more.
Do I need a degree?
DataAnnotation’s baseline for general work is a bachelor’s degree or equivalent real-world experience. Specialist tracks require advanced degrees or credentials. Outlier’s requirements vary by project.
Can I retake the DataAnnotation assessment?
No. The platform states you can take the Starter Assessment only once, so prepare before you begin.
How long until I get my first task?
Anywhere from a few days to several weeks, and sometimes never if you are not accepted or no matching project is open. Plan for patience.
Can I do this from Pakistan?
It depends on the platform and project. PayPal is not available in Pakistan, which affects DataAnnotation’s payout method. Outlier lists Airtm and bank transfer alongside PayPal, but it does not publish a full country list, so check availability and payout options inside your own account before investing time. Never use a VPN or someone else’s details to get around restrictions.
Can I use AI tools during assessments?
Assume no unless the instructions clearly say otherwise. Submitting AI-written answers can get you rejected or banned.
How much can I realistically earn?
It varies widely. Beginners on general tasks often see modest weekly income, while verified specialists can earn more. Use the planning model above, and expect gaps between projects.
Is the work stable enough to replace a job?
For most people, no. Task volume depends on client demand, so treat it as flexible side income alongside other skills such as freelancing, content or tutoring.
Final thoughts
AI training is one of the few online earning paths where your judgment, writing and subject knowledge are the product. It rewards careful people, and it punishes rushing and shortcuts. Prepare your documents, choose one track, apply once and do it well, set up a payout method that actually works in your country, and keep a second income stream running in parallel. If the platform has work, you are paid for your thinking. If it does not, you have lost a few hours, not your savings.
Get new earning guides on WhatsApp
Sources and further reading
- DataAnnotation FAQs (application process, pay tracks, PayPal payments, identity verification)
- Outlier: payment support update (supported payment providers)
- Outlier official site
- Independent contributor reviews and comparisons (2026) used for pay ranges, acceptance-rate reports and work-availability patterns. These are self-reported and change often.
Disclaimer: This article is for general information, not financial or legal advice. Pay rates, eligible countries, payout methods and platform rules change often and vary by project. Earnings are not guaranteed. EarnGuide has no affiliate relationship with the platforms mentioned in this post.
Online earning tips, passive income ideas and money-making strategies for beginners.
www.earnguide.online
