Data annotation jobs are paid work labeling, rating, and correcting the data that AI models learn from. 39 live postings in the catalogue carry an annotation, labeling, AI trainer, AI tutor, or rater title, and 36 of them are remote. Pay, employers, and Levels below all come from those postings.
The number this article rests on: 39 live data annotation postings, 28 of them with a published hourly USD rate. Measured September 2026.
This page counts by job title, with a strict filter: a posting only enters the set when its title names annotation, labeling, rating, AI training, or model evaluation. The AI training jobs guide takes a wider view of the same field, so read the two together rather than as rival counts. The AI tutor jobs guide covers teaching roles, a neighbouring kind of remote work.
How much do data annotation jobs pay?
Data annotation postings that publish an hourly USD rate have a median floor of 35 US dollars an hour and a median ceiling of 45 US dollars an hour, across 28 listings. Those are the employers' own published ranges, read from the posting and never estimated. Measured September 2026.
Only 4 of these postings print an annual salary instead, so no annual median is quoted. Hourly and annual figures are kept apart here and never converted into each other, because a rate says nothing about how many hours the work offers.
Contractor platforms that recruit annotators directly publish their own rate cards outside job boards. What one of them states about its pay, and how to check it, is covered in is data annotation legit.
Who is hiring for data annotation work?
xAI posts the most data annotation roles in the catalogue, with 28 of the 39 live listings, or 71.8%. The employers with the most openings are xAI (28), Cohere (4), and Anthropic (1). A set this concentrated moves sharply when one employer opens or closes a batch of roles. Measured September 2026.
DataAnnotation and Outlier, the two platform names searched most alongside this keyword, have no live postings in the catalogue under their own names. Both recruit through an assessment on their own sites, as their public FAQ pages describe, so a job board count understates how much work they offer. The Outlier review sets out what that platform publishes about itself. Scale AI, the company Outlier names as its operator, has 222 live postings of its own across all job titles.
Are data annotation jobs remote?
36 of the 39 data annotation postings say remote, or 92.3%, against 48.8% across the whole catalogue. That puts annotation work above the rest of the market on remote work, for a job that is done in a browser. Measured September 2026.
Remote status is read from the posting text. A role that describes an office without using the word remote is counted as on-site, even when the tasks themselves could be done from anywhere, so the real share is at least this high.
What does a data annotation job involve?
4 of the 39 postings score Level 1, 0 Level 2, 31 Level 3, and 4 Level 4, so the largest group is Level 3. The Level says how central AI is to the role, from a tool in the workflow to the product itself. Measured September 2026.
By category, the postings sit mainly in education (26), data (5), and research (3), a sign that some annotation roles are hired for subject knowledge rather than generic data work. The tools named most often are Runway, Sora, and After Effects, which suggests that part of the work involves generated images or video.
Across these titles the core task is the same: read a task, apply written guidelines, and label, rank, or rewrite a model's output so the next version learns from it. The Levels page explains how the scale separates using an AI tool from working on the model itself.
Are data annotation jobs the same as AI training jobs?
Mostly, though the counts differ. A broad title filter that also catches trainer and labeling words used outside annotation finds 14 live postings; the strict filter behind this page finds 39. The gap is mostly roles that only borrow the vocabulary, such as training roles unrelated to AI models. Measured September 2026.
Use the strict count when you want annotation work itself, and the broad one when you are open to any role that trains or evaluates models. Neither count includes tasks offered only inside a platform account, because those are never published as postings. For a first step into AI work with a clearer career ladder, the AI internships guide covers student roles across every category.
The jobs right now
Live postings that mention annotation, newest first, generated from the catalogue rather than written by hand.
