What is data annotation?

Data annotation is the human work of labeling raw data — text, images, audio, video, or code — so that machine learning models can learn patterns from it. Every large AI model you've used, from ChatGPT to self-driving cars, was trained on millions of examples annotated by people.

The four main types

• Text annotation: rating AI responses, writing model answers, checking factual accuracy, ranking two chatbot outputs.
• Image annotation: drawing bounding boxes around cars, tagging objects, segmenting medical scans.
• Audio annotation: transcribing speech, labeling emotions, marking speaker changes.
• Code annotation: reviewing AI-generated code, writing test cases, fixing bugs the model produced.

What does it pay?

Rates vary widely by task and expertise. Generalist text work on platforms like Outlier or DataAnnotation typically pays $15–$25/hour. Specialist work — PhD-level physics, legal review, medical coding, senior software engineering — can pay $40–$150/hour. Rates almost always scale with domain expertise and the difficulty of the task.

Who hires annotators?

The big players are Scale AI, Outlier, Surge AI, Mercor, DataAnnotation.tech, Invisible, and Labelbox. Frontier labs (OpenAI, Anthropic, Google) usually contract through these platforms rather than hiring annotators directly.

Is it a real job?

Yes — and it's growing fast. See our guide on whether data annotation is legit for pay reports and platform reviews, or jump to how to get started.