What annotation work actually involves
Annotation is the labeling side of AI work: you tag, sort, transcribe or judge raw data so a model can learn from it. It overlaps with AI training work, but the tasks are shorter, more rule-bound and more often visual or audio. These are the task types that appear across platforms today:
| Task type | What you do | Where it appears |
|---|---|---|
| Image labeling | Draw boxes or outlines around objects, tag image content | Remotasks ("label images") |
| Audio transcription | Type what is said in a clip, mark speakers and noise | Remotasks ("transcribe audio"), CrowdGen linguist work |
| LiDAR and sensor data | Label 3D point clouds and multi-camera scenes for robots and vehicles | Appen's physical AI data products |
| Search and ads relevance | Judge whether a result or ad matches a query in your language | TELUS Digital ads assessor roles, Mindrift search tasks |
| Content safety | Flag harmful or policy-breaking content | TELUS Digital internet safety evaluator roles |
| Text classification | Sort AI drafts or user text by tone, clarity or category | Mindrift "label and classify" tasks |
| Video capture | Record everyday or work activities on a phone | Mindrift video project at $5 to $15 per hour |
Checked 17 October 2026 at each platform's own page.
Appen's site describes LiDAR point cloud annotation and multi-camera sensor fusion as part of its physical AI work, a reminder that AI annotation jobs now include robotics data, not just photos. If you would rather write prompts and rank chatbot answers, that is closer to AI training jobs that pay for writing and rating, a separate guide with its own platform grid.
Data annotation platforms compared
These seven data annotation platforms publish enough about entry, pay and payout to compare. Each is remote; the difference is how you get in.
| Platform | Run by | Work type | Published pay | Payout | Best for |
|---|---|---|---|---|---|
| Mindrift | Toloka | Rating, classifying, search relevance, video capture | $15 to $30/hr entry level | Twice a month, Payoneer or Tipalti (PayPal) | Beginners with no test history |
| Remotasks | Scale AI | Image labeling, audio transcription, robotics annotation | Pay rises with task difficulty | Weekly, PayPal or AirTM | Labeling beginners outside the US |
| TELUS Digital AI Community | TELUS Digital | Ads and search assessors, data analysts, safety evaluators | Set per role and country | Agreed at onboarding | Native speakers of a listed language |
| CrowdGen | Appen | Linguistics, transcription, domain annotation | Up to $95/hr for linguists | Bank transfer, PayPal, Payoneer, Airtm | Linguists and specialists |
| Alignerr | Labelbox | Labeling and rating on the Labelbox platform | $20 to $40/hr Arabic localization; up to $150/hr computer vision experts | Stripe Connect or PayPal | Computer vision and language experts |
| DataAnnotation | DataAnnotation | Labeling, image testing, chatbot rating | $20+/hr multilingual; $25 to $50+/hr general | PayPal, a few days after request | Careful readers who can pass one test |
| Outlier | Scale AI | Rating and ranking model outputs | Rate shown before each project | Weekly on Tuesdays | Graduates with a degree |
Checked 17 October 2026 at each platform's own page.
Verdicts in brief. Mindrift is the easiest door: free to join, with tasks you can start right away and qualifications that open better-paid specialized projects later. Remotasks is the strongest pure labeling platform for international workers, with weekly payouts and training built in. TELUS Digital is the only one here that hires through dated job postings with a formal application, which suits people who want a structured part-time role. CrowdGen and Alignerr pay most for specialists. DataAnnotation and Outlier lean toward writing and reasoning; read our DataAnnotation platform review and our review of Outlier before applying to either.
Entry tests: what each platform asks
The entry test decides more than your resume does. Here is what each platform says it uses.
| Platform | Entry test | Time or rule | Other checks |
|---|---|---|---|
| DataAnnotation | Starter Assessment, plus optional specialist tests | About 1 hour; specialist tests 1 to 2 hours; one attempt only | Persona ID check |
| Outlier | Skill screenings during onboarding | 30 to 90 minutes in total | ID and phone from your country |
| TELUS Digital AI Community | Language proficiency, critical thinking and role-specific assessments | Part of a five-step hiring process | CV and IP address validation, ID upload |
| Mindrift | Short qualification tasks per project | All-access tasks open before any test | Free to join |
| Alignerr | Domain assessment or AI interview, for some jobs | You are notified if a job requires one | Phone and Persona ID check |
| Remotasks | Online courses and hands-on training | Harder tasks open as you train | Weekly payout setup |
| CrowdGen | Project qualification | Most projects have one | Never asks you to pay |
Checked 17 October 2026 at each platform's own page.
TELUS Digital's AI Community hiring page is the most explicit about compliance: after the assessments it validates your CV, checks that your IP address matches the location you applied from, and confirms you are not on a conflicting project. Use your real home connection, without a VPN, for any test on any platform. DataAnnotation's FAQ makes the same point about VPNs during its ID check.
How to pass an annotation entry test
- Read the guidelines twice before starting (20 to 30 minutes). Annotation tests grade rule-following. Most failures come from missing an edge case the instructions already covered.
- Work slowly on the first ten items. DataAnnotation says it reviews for thoroughness and accuracy rather than speed. Speed comes later, once you are on paid tasks.
- Write out your reasoning when a box is offered. Several platforms ask you to justify a judgment. Short, specific reasons score better than long general ones.
- Check location and language settings. TELUS Digital roles are tied to one country and one language, such as Japanese in Tokyo or Portuguese in Brazil.
- Keep your ID and phone ready. Alignerr starts with a text message to your phone; DataAnnotation asks for front and back photos of a government ID plus a selfie.
What data annotation jobs pay
Annotation pays less than expert AI training work, and the spread depends on language and domain skills. Every figure here is as each platform publishes it.
| Work | Platform | Published pay | Unit |
|---|---|---|---|
| Video capture of everyday activities | Mindrift | $5 to $15 | per hour |
| Entry-level tasks | Mindrift | $15 to $30 | per hour |
| Multilingual annotation | DataAnnotation | $20+ | per hour |
| Arabic localization | Alignerr | $20 to $40 | per hour |
| General projects, including labeling | DataAnnotation | $25 to $50+ | per hour |
| Linguists | CrowdGen | Up to $95 | per hour |
| Computer vision and ML experts | Alignerr | Up to $150 | per hour |
Checked 17 October 2026 at each platform's own page.
Two patterns stand out. Language work is the best-paid entry point for people without a degree, because native fluency is the qualification. And computer vision experience is the main way to move from generalist labeling rates into three-figure hourly rates. Mindrift adds a condition worth knowing: you are paid only for tasks that are completed and accepted, so rejected work earns nothing.
The Mechanical Turk closure
For years, Amazon Mechanical Turk was where many people did their first labeling tasks. Its worker page now states that Mechanical Turk permanently closed on 30 September 2026. If you relied on it, the closest replacements by task type are Mindrift's all-access tasks for short ratings, Remotasks for image and audio labeling, and TELUS Digital for structured part-time rating roles.
Before you sign up: checklist
- Joining is free. Mindrift says joining is always free, and CrowdGen says it will never ask you to pay to join a project.
- The domain is official. Appen lists crowdgen.com and app.crowdgen.com among its legitimate sites.
- The ID check runs through a named provider. DataAnnotation and Alignerr both use Persona.
- Pay is shown before the task. Mindrift says every task shows a fixed reward before you start.
- Payout methods are listed. Expect PayPal, Payoneer, Airtm, Stripe Connect or bank transfer, plus a schedule.
- You can start part time. Most TELUS Digital AI Community postings are listed as part time and remote.
Scam warning
Scammers copy annotation language: "online assessor," "data labeling," "product boosting." The FTC's April 2026 alert on fake job texts describes recruiters offering "online assessor" work by text and asking you to reply "YES," then pushing a fake check or a task scheme that ends with you depositing money. No real annotation platform contacts you first by text, WhatsApp or Telegram, and none asks for a deposit. Apply only through the platform's own domain.
How we researched this
On 17 October 2026 we read the home, FAQ, process and job listing pages for each platform in this guide, the TELUS Digital AI Community listings and hiring process, Appen's data product and scam awareness pages, the Amazon Mechanical Turk worker page and the FTC's 2026 job text alert. We recorded each platform's entry test, published pay with its unit, payout method and schedule as stated on its own pages, then compared them on those fields using HumanRank's ranking method. For a wider view of the market, including salaried roles, see the AI jobs hub; students should also check our AI internships list, and career changers our guide to no-experience AI roles.
Frequently asked questions
What are data annotation jobs? They are paid tasks that label raw data for AI: drawing boxes on images, transcribing audio, rating search results or classifying text. Most are remote contract work paid per hour or per task.
How much do data annotation jobs pay? Published entry rates run from $15 to $30 per hour at Mindrift and $25 to $50+ per hour for DataAnnotation general projects. Specialist work pays more, such as up to $150 per hour for computer vision experts at Alignerr.
What is the easiest data annotation platform to join? Mindrift, by its own description: it is free to join and you can start on all-access tasks right away, before any qualification test.
What is on a data annotation entry test? Usually a short set of real tasks graded against written guidelines. TELUS Digital adds language proficiency and critical thinking assessments, and DataAnnotation's Starter Assessment takes about an hour.
Can I retake a failed annotation test? On DataAnnotation, no: the Starter Assessment allows one attempt. Other platforms set rules per project, so read the instructions before you begin.
Do AI annotation jobs need a degree? Often not. Mindrift's all-access tasks and Remotasks' training are open to beginners, while Outlier asks for at least an associate degree.
Is Amazon Mechanical Turk still open? No. Its worker page says Mechanical Turk permanently closed on 30 September 2026.
Why do annotation platforms check my IP address? To confirm you work from the country you applied from. TELUS Digital lists IP address validation as part of its compliance checks, so turn off any VPN.