Getting started
Data annotation jobs: pay, skills and scams to avoid
3 min read · Updated
Data annotation, also called data labeling, is the work of adding meaning to raw data so machine-learning systems can learn from it. It is one of the most accessible ways to start earning in the AI industry, and one of the most scam-prone. Here is a clear-eyed look at the work, the pay and how to protect yourself.
What annotators do
- Text: classify sentiment, tag names and places, check translations, rate answer quality.
- Images and video: draw boxes, outline objects, label scenes, mark actions or describe what is visible.
- Audio: transcribe speech, mark speakers, rate pronunciation or flag background noise.
- Search and ads: judge whether results are relevant to a query.
- Specialist data: label medical scans, legal clauses or code snippets, usually requiring credentials.
Most tasks come with a guideline that defines each label precisely. Your job is to apply the rules consistently, not to invent your own.
Skills that matter
You do not need a degree for basic annotation, but you do need reliability. Platforms measure accuracy against reference answers and against other annotators. Strong candidates read instructions carefully, notice small differences and stay focused for long sessions.
- Good reading comprehension in the project language.
- Comfort learning simple tools quickly.
- Patience with repetitive work and accuracy under mild time pressure.
- Subject knowledge for specialist projects: medicine, law, coding, mathematics or a native-level second language.
What pay looks like
Rates vary by platform, task, language and country, and they change often, so treat any figure as indicative. General-purpose annotation tends to pay modestly per hour or per task. Work that requires a specialist qualification typically pays much more. Per-task pay can look generous until you divide by the time the task really takes.
Compare opportunities by estimated effective hourly pay: reward per task divided by the average minutes it takes you, including reading the guideline and any unpaid qualification time. Also ask whether pay is guaranteed for approved work or can be rejected after the fact.
How to recognise a legitimate platform
- It has a real company website, a named legal entity, a privacy policy and terms of service.
- It pays workers; it never charges you to apply, train or "activate" an account.
- Payment methods, pay schedule and minimum payout are written down.
- Reviews from independent workers exist, including critical ones, and the platform responds to support emails.
- Identity verification is done through its own site, not through a chat message.
On Chakri.ai every global listing shows the source platform and a safety indicator, and we hide listings that fail our automated scam checks. We earn a commission when you apply through some of our links, which never changes what you pay. Read the affiliate disclosure for details.
Common scams
- Advance-fee scams: you are told to pay for training materials, a starter kit or a license before you can work.
- Check scams: you receive a payment, are asked to keep part of it and send the rest elsewhere. The original payment later bounces.
- Fake recruiters: someone impersonates a known company on social media or messaging apps and moves you off-platform quickly.
- Data-harvesting forms: a long "application" that mostly collects ID documents, bank details and selfies with no clear employer.
- Task-reward schemes: you are paid small amounts to complete "tasks" and then pressured to deposit money to unlock larger payouts.
If something goes wrong, stop contact, keep screenshots, tell your bank or payment provider and report the listing to the platform where you saw it.
Protecting your data
- Share ID documents only through the official verification flow on the platform domain.
- Use a unique password and two-factor authentication.
- Do not give out one-time codes or install remote-access tools.
- Never annotate data that includes real personal information unless the project is clearly authorised and you have signed the right agreements.
Building a sustainable routine
Treat annotation as a skill that compounds. Track your accuracy, ask for feedback, take on harder task types as you qualify and keep notes of guideline changes. Spread across a few platforms to smooth the gaps between projects, and take breaks: long stretches of repetitive work lower quality.
Ready to look for work?
Browse live AI training, annotation and RLHF openings, each labeled with its source platform.
Browse jobs