AI Data Annotation & Model Rating · Unit 8 of 8
Unit 8: The job: platforms, money, scams and your profile
How remote AI data work really works, contractor money basics, how to spot and report job scams, and how to build a profile, a portfolio and your next step.
Opens the first lesson of this unit in AI Data Annotation & Model Rating. Every lesson is open from the start.
Lesson 1: How remote AI data work really works
Goal: Describe the usual path from application to paid tasks, and plan for variable work and pay.
Most remote AI data work is contract work found through platforms and vendors. AI companies hire data companies; the data companies recruit contributors like you through a platform, usually as independent contractors rather than employees. Some roles, especially reviewers and team leads, are offered as fixed-term or salaried jobs, but at entry level expect contract terms. The usual path looks like this:
- Apply on the platform's own website, with your location, languages, education and skills.
- Pass a screening assessment: a general test of reading, writing, reasoning or a short rating task.
- Pass qualification tests for particular projects or skill areas. Many assessments are unpaid, so read the terms first.
- Onboard onto a project: study its guideline, sometimes with training tasks.
- Work in the task queue, judged by gold questions and audits (Unit 4).
- Get paid on the platform's schedule, by the method it offers.
Task flow is variable. Projects start and end suddenly, queues empty without warning ("no tasks available"), and your access can change with client demand or with your quality scores. Many experienced contributors keep accounts on more than one platform and never rely on one project for essential bills. Pay varies widely by project, skill, language and country, and it changes over time, so this course states no rates. Read the current terms on the platform itself, and treat any rate quoted online as one person's experience at one moment.
| Kind of place | Examples (as of November 2026) | Notes |
|---|---|---|
| Platforms recruiting AI trainers and raters | DataAnnotation; Outlier (run by Scale AI) | Writing, rating and coding tasks for generalists and experts. Work available changes with client demand. |
| Long-running crowd and language-data vendors | CrowdGen by Appen (formerly Appen Connect); OneForma (Centific); Mindrift (Toloka) | Rating, transcription, language and annotation projects. |
| A reminder that platforms change | Amazon Mechanical Turk | One of the oldest crowd-work marketplaces. Amazon closed it on 30 September 2026. |
Worked example: reading a project's terms before you start
- "Assessment: about one hour, unpaid." → You decide whether the time is worth it.
- "Pay is shown in your dashboard and varies by project." → You will see the real rate before you work, not in an advert.
- "Tasks are not guaranteed, and projects may pause without notice." → Plan your budget for variable income.
- "You are an independent contractor and are responsible for your own taxes." → Lesson 2 shows what that means.
- "We will never ask you to pay a fee." → A good sign, and true of every legitimate platform.
Five lines of terms tell you the time cost, how pay works, the risk to your income, your tax duty and a scam safeguard.
- Independent contractor
- Someone paid for work as a separate business rather than as an employee, usually without benefits or tax withheld.
- Screening assessment
- A general test a platform uses to decide whether to accept a new contributor.
- Task queue
- The list of tasks available to you on a project. It can empty without warning.
- Data vendor
- A company that supplies labelled data and human evaluation to AI companies, often through a contributor platform.
- Variable task flow
- Work that rises and falls with client demand, so hours and income change from week to week.
Remember: Apply on official sites, take assessments seriously, read every term, and never depend on a single project.
Lesson 2: Contractor money basics
Goal: Explain in general terms how taxes, records and budgeting work for contract income, and where to check the rules.
As an independent contractor, nobody usually withholds tax from your pay, and you normally get no employer benefits such as paid leave, sick pay, health insurance or retirement contributions. That changes how you should handle every payment. This lesson is general information, not tax or legal advice: rules depend on your country and situation, so check your tax authority's official website or ask a qualified professional.
In the United States, the IRS says self-employed people generally owe self-employment tax (Social Security and Medicare) on top of income tax, and gives its rate as 15.3%, owed once net earnings from self-employment reach 400 dollars in a year. Because no one withholds it, you may need to make estimated tax payments during the year, usually quarterly. A platform may send you a form such as 1099-NEC reporting what it paid you, but you must report all of your income whether or not a form arrives. Other countries have their own systems, often involving registering as self-employed and filing a yearly return.
- Set aside a share of every payment for tax in a separate account, using a percentage based on official guidance or a professional's advice for your situation.
- Keep records of every payment and platform statement, and receipts for work expenses you may be allowed to deduct, such as part of your internet bill. Ask what qualifies where you live.
- Budget for variable income: base essential spending on your lowest typical month, not your best one, and build a small buffer for weeks without tasks.
- Read how and when you are paid, including any minimum payout and the payment method, and keep your payout details secure.
Worked example: one payment, handled well
A payment of 500 dollars arrives from a platform. (An example number for the arithmetic, not a typical amount.)
- You move a set share, say 25%, into a separate tax account: 500 × 0.25 = 125 dollars. Your right percentage depends on your situation; this only shows the habit.
- You record the date, the amount and the platform in a simple spreadsheet.
- You save the platform's statement and any receipts for work expenses.
- When an estimated payment is due, if your situation requires one, the money is already waiting.
The tax share was never yours to spend. Setting it aside at once avoids a painful bill later.
- Self-employment tax
- In the US, the Social Security and Medicare tax that self-employed people pay on their net earnings.
- Estimated tax payments
- Tax paid during the year, often quarterly, by people whose income has no tax withheld.
- Form 1099-NEC
- A US form that some clients send contractors to report what they paid. All income must be reported whether or not one arrives.
- Variable income
- Income that changes from week to week or month to month.
Remember: No one withholds your tax: set money aside from every payment, keep records, budget on your lowest month, and check official guidance.
Lesson 3: Job-scam safety
Goal: Recognise the red flags of fake AI-training jobs, check a job safely, and report scams.
Because AI training jobs are remote, popular and found online, scammers copy them. They post fake listings, message people out of the blue and pretend to be real platforms. The most important rule comes straight from the US Federal Trade Commission (FTC): "Never pay to get paid or get a job." Legitimate work never asks you to pay to start, to buy equipment, or to deposit a check and send money on.
| Red flag | Why it means a scam |
|---|---|
| You must pay to start: a training fee, a "starter kit", software, a certificate or a deposit | Real platforms pay you. They do not charge you to work, and "refundable" fees are a common trick. |
| They send you a check to buy equipment, then ask you to send part of it back or on to their "supplier" | The check bounces days later, after your real money has gone. A deposit can look cleared and still bounce. |
| "Task" jobs where you must add your own money, often cryptocurrency, to unlock tasks or withdraw earnings | The FTC warns about these task scams. The earnings shown on screen are fake. |
| Unexpected texts or WhatsApp or Telegram messages offering work, and interviews held only by chat | The FTC says real employers will not recruit you that way. |
| Pay far above normal for little work, often quoted per day | If it sounds too good to be true, it is. |
| Requests for your bank login, card number or ID before any real offer | This is identity theft. Legitimate platforms collect tax and payout details only after you join, inside their own site. |
| An email address or website that is not the company's real one | Impostors use look-alike addresses and free email accounts. |
How to check a job. Find the company's official website yourself, without using any link in the message, and apply only there. Confirm the job appears on its official careers or join page. Search the company's name with words such as "scam" or "complaints". Remember what normal looks like: legitimate platforms may give you an unpaid qualification test, and may ask for tax and payout details after you join, inside their own site. They will not ask you for money. In the US, report job scams at ReportFraud.ftc.gov; elsewhere, report them to your national consumer or fraud agency.
Worked example: spotting a fake
A chat-app message: "Hi! Our AI training team found your CV. Earn 300 to 800 a day rating apps, no experience needed. Pay a small refundable activation fee to unlock your task account."
- An unexpected message on a chat app: red flag.
- "Found your CV", with no named company or official website: red flag.
- Daily pay far above normal, with "no experience needed": red flag.
- A fee, "refundable" or not: the decisive red flag. Real platforms never charge you to work.
- You do not reply. You block the sender and report the message.
One red flag is enough to stop. This message has four.
Hands-on lab in the room: Scam-or-legit job sorter. Sort eight job posts and messages into legitimate or scam, and name the main red flag in each scam.
- Advance-fee scam
- Any demand that you pay money up front to get work or to receive money. A sure sign of a scam.
- Fake check scam
- A scam where you are sent a check, asked to pass some of the money on, and left owing the bank when the check bounces.
- Task scam
- A fake job where you "earn" money for simple online tasks but must deposit your own money to unlock tasks or withdraw.
- Impostor scam
- A scam where criminals pretend to be a real company or recruiter.
Remember: Never pay to get a job. Never send money on from a check. Apply only on official sites, and report scams.
Lesson 4: Your profile, portfolio and next step
Goal: Write a specific, truthful profile, build a portfolio from your own material, and name the next role to aim for.
Platforms match people to projects using their profiles and test results, so both deserve care. Your profile should state your languages and honest levels, your education, and every area where you have real expertise: a degree subject, a profession such as nursing, accounting, teaching or a trade, a programming language, or deep knowledge of a field. Domain skill is what moves people from general tasks to specialist work. Be accurate, because projects test claimed skills, and overstating them leads to failed qualifications or removal.
Assessments are your audition. Take them when you are rested and have uninterrupted time, read every instruction twice, follow any style guide exactly, and check your work before submitting. Many platforms allow no retake, or only after a long wait.
- This course's final project: a guideline you wrote, labelled examples, an agreement check with kappa, side-by-side ratings with rationales and a fact-checked ideal response.
- Two or three writing samples: an explanation, a rewrite of a weak answer, a careful fact-check.
- If you code, small, well-tested code examples in a public repository, with notes on edge cases.
- On a CV or professional profile, skills with evidence: "Labelled 200 practice items against a four-page guideline, kappa 0.82 against a second rater" says far more than "detail-oriented".
Build the portfolio from your own or public material only, never from task data. Your next step to aim for: pass the qualification tests for one generalist AI trainer or rater project and one project in your strongest area, then build a quality record. With a strong record, the usual next roles are reviewer or QA (checking other people's work), then specialist or domain-expert trainer, and later annotation team lead, data quality analyst or trust and safety analyst. On this site, Prompt Engineering & AI Automation and Machine Learning Fundamentals are natural next courses, and Data Lab helps with quality and analyst work.
Worked example: turning experience into a profile line
- Experience: five years as a pharmacy technician.
- Weak line: "Medical knowledge."
- Strong line: "Five years as a pharmacy technician: medicine names, dosing labels and patient instructions. Can check health answers for accuracy and safe wording."
- Evidence from this course: "Completed a 32-lesson course in AI data annotation and model rating, including fact-checking, safety labelling and inter-annotator agreement."
Specific, checkable, truthful, and pointed at the projects you want.
- Contributor profile
- Your record on a platform: languages, skills, education and expertise, used to match you to projects.
- Portfolio
- A collection of your own work samples that shows what you can do.
- Quality record
- Your history of gold accuracy, audit results and feedback on a platform.
- Data quality analyst
- A role that measures and improves the quality of labelled data, using agreement, audits and error analysis.
- Trust and safety analyst
- A role that applies and improves content policies and investigates harmful content or behaviour.
Remember: Be specific and truthful on your profile, treat every assessment as an audition, build a portfolio from your own material, and aim for reviewer, then specialist.
Practise it in AI Data Annotation & Model Rating
Each lesson ends with three questions that check you can use the idea, not just recognise it. Wrong answers explain why and come back in Review. The unit ends with a 7-question quiz; pass it with 80% or more to master the unit. Hands-on lab: Scam-or-legit job sorter.
Every lesson is free: no ads, no lives and no paywall.