AI models learn from examples, and many of those examples are created, labelled or checked by people. That has created a large market for online work with titles such as AI trainer, data annotator, rater, evaluator or prompt writer. Some roles require only careful reading and good language skills; others look for experts in coding, mathematics, medicine, law or specific languages. The work is often remote and flexible, but it is also an area where scams have grown fast.
This guide explains what the work involves, how to judge an offer and how to protect yourself. We do not list vacancies, recommend platforms or quote pay.
What AI training work involves
Typical tasks include:
- Labelling data: tagging images, audio or text with categories.
- Rating responses: comparing two answers from an AI system and choosing the better one according to guidelines.
- Writing examples: creating questions and high-quality answers in a subject you know well.
- Checking facts and safety: flagging incorrect, biased or harmful outputs.
- Transcription and translation: producing accurate text in one or more languages.
Most projects come with detailed guidelines and quality checks. Workers are usually tested before getting access and are monitored for accuracy.
Skills platforms look for
- Excellent reading and writing in the project language.
- Attention to detail and patience with long guidelines.
- Subject expertise for specialist projects, such as programming or healthcare.
- Ability to work independently and meet deadlines.
- A reliable computer and internet connection.
How the work is usually organised
- Freelance or contractor status is common. You may be paid per task or per hour and be responsible for your own taxes.
- Employment contracts exist with some companies and outsourcing firms.
- Availability varies. Projects can start and end suddenly, so income may be irregular.
- Location rules apply. Platforms often accept workers only from certain countries.
Read the terms carefully: how payment is calculated, when and how you are paid, how quality is assessed and what happens if you are removed from a project.
The regulatory context
Rules on AI are developing. In the EU, the Artificial Intelligence Act (Regulation (EU) 2024/1689) sets obligations for providers of AI systems, including requirements related to data quality for certain high-risk systems. You do not need to know the regulation in detail to do annotation work, but it explains why quality guidelines are so strict.
Task scams: the biggest risk
Fake “task” jobs are widespread. They often look like AI training or app-rating work. Typical pattern:
- You are contacted out of the blue by text, WhatsApp, Telegram or social media.
- You are offered easy online “tasks”, such as rating products or apps, and shown small early earnings.
- You are then asked to deposit money, often in crypto, to “unlock” more tasks or withdraw your earnings.
- The earnings never arrive.
Genuine platforms never ask workers to pay to work or to withdraw earnings. Other warning signs:
- No clear company name, address or contract.
- Pay that seems far above similar work.
- Pressure to recruit friends.
- Communication only through messaging apps.
Report fraud at ReportFraud.ftc.gov in the US or to your national authority.
How to find genuine work
- Search for the company directly, not through links in messages.
- Check the legal entity, its address and its terms of service.
- Look for clear assessments: genuine platforms usually test applicants.
- Use official job portals where possible, such as CareerOneStop in the US or EURES in Europe.
- Protect your data: do not send identity documents until you have verified the company and understand why they are needed.
Official US government site. We are not an employer or platform.
A typical working session
Most annotation work happens in a web-based tool. You log in, choose or receive a batch of tasks, read the guidelines for that project and work through items one by one. Quality reviewers check a sample of your work, and your access may depend on keeping accuracy high. Good habits include re-reading guidelines when they change, keeping notes on tricky cases, taking regular breaks to stay focused and never using AI tools to complete tasks unless the project explicitly allows it.
Building a profile
- Highlight languages, writing samples and subject expertise.
- Mention experience with guidelines-based work, such as transcription, moderation or quality assurance.
- Keep a record of projects completed and quality scores, where allowed.
Frequently asked questions
Do I need a degree to train AI?
Not for general annotation. Specialist projects may require expertise or qualifications.
Is AI training work stable?
It can be irregular. Treat it as flexible work unless you have an employment contract.
Is it normal to pay a deposit to start?
No. A request for payment is a strong sign of a scam.
Can I do this from any country?
Platforms usually limit eligibility by country. Check the terms.
Sources
Related guides
- Supermarket Bagger and Courtesy Clerk Jobs: A Good First Job Explained
- Cleaner and Janitor Jobs: Duties, Chemical Safety and How to Get Hired
- Local Government Jobs: How Public Sector Hiring Works and How to Apply
- Packing and Packaging Jobs: What Packers Do, Safety Rules and How to Apply
- Shelf Stocker and Merchandiser Jobs: Duties, Night Shifts and How to Apply
Opportuly Careers is an independent guide. We do not hire, represent any platform or charge for information.
