The basics
Annotation, evaluation, training: what the words mean
Listings use these words loosely. Here is what each one usually means for the work you would do.
Annotation
You label or mark up data: tagging parts of an image, transcribing audio, classifying text, marking where an answer goes wrong. The task is defined by a guideline, and quality is checked against it. Some annotation is quick and repetitive. Expert annotation, such as marking errors in a legal or medical answer, is slower and pays more.
Evaluation
You judge model output. That can mean comparing two answers, scoring one against a rubric, or explaining exactly why a response is wrong. Evaluation roles usually ask for written reasons, not just a score, so clear writing matters as much as subject knowledge.
“AI training”
Platforms use this as an umbrella for annotation, evaluation, writing example answers and creating hard prompts. The phrase does not mean you build or code the model. Read the task description, not the title, to see which kind of work it is.
Data collection and studies
Some projects pay you to record your voice, take photos or video, complete a survey, or join a research interview. These are short and usually need no specialist knowledge, but they often have strict eligibility by country, device or demographic.
Why the difference matters
The work type tells you what the assessment will test, how pay is likely to be set (per hour, per task or per study) and whether your profession gives you an edge. On Remotefolio, the work type filter on the role list uses these groups.
These guides are general information, not legal, tax or career advice. Check each platform’s own terms before you apply.