People searching for AI training sites that pay need two different checks: whether a platform describes work they could do, and whether a particular offer has acceptable eligibility and payment terms. A company name, an advertised hourly rate or a list of platforms cannot answer both questions.
Use the official-source comparison below to build a shortlist, then complete a separate offer check before committing time or sharing documents. These are examples to investigate, not a ranking, a report of earnings from tested accounts or confirmation of current openings in your country.
Published by Spirelight, which also provides its own contributor signup. First published . The external descriptions below were checked on 18 October 2026; platform requirements can change.
Compare what the official platform actually describes
Start from each provider's official website and follow its own application links. A platform description establishes what the provider says about its service; it does not authenticate a separate recruiter message, prove reliable payment on your assignment or establish that you qualify.
| Platform and source | Work described | Question for your shortlist |
|---|---|---|
| DataAnnotation FAQ | Chatbot-response evaluation, writing, coding and specialist work; requirements differ by track. | Which track fits your actual skills, and what do its current application, qualification and payment rules require? |
| Outlier FAQ | Expert contributions such as generating prompts, ranking outputs and improving responses; participation is as an independent contractor. | Do you meet the specific opportunity's expertise and location requirements, and what project-specific onboarding applies? |
| RWS TrainAI community | Examples include rating, data collection and annotation across text, audio, images and video. | Which advertised role matches your skills and equipment, and what does its particular brief require? |
| Spirelight participation guide | Speech recording, transcription, transcript checking and audio annotation projects. | Does speech-data work fit you, and can an accurate profile be considered when a suitable project becomes available? |
Keep the descriptions separate from the conditions of an actual offer. For example, DataAnnotation's FAQ describes hourly project compensation, so an hourly listing is not itself a scam test. Outlier describes both general and project-specific onboarding. RWS lists several different task types. None of those descriptions supplies every condition of the assignment you may be considering.
Make one comparison card for each actual offer
Record the official project link and the date you checked it, then answer these questions. Mark an unknown as a question to resolve rather than treating a missing detail as a promise.
- Eligibility and deliverable: What location, age, language, expertise and equipment conditions apply? What exactly would you submit, and what counts as acceptable work?
- Application effort: Which assessments or qualifications are required? What time might they take, and does the specific agreement say whether any of that time is paid?
- Payment unit: Is the quoted amount for an item, a batch, eligible working time or a whole project? Which activities and review outcomes make it payable?
- Access to the money: What currency, payout method, schedule, threshold, transfer fees and conversion conditions apply to your account and country?
- A problem later: Which independently verified official channel handles an unclear review or an unpaid agreed balance? Keep the applicable brief, submission and payment records.
If the work includes recordings, read the agreement's intended uses, recipients, retention and sharing conditions. Check what it says about speech synthesis or voice cloning before recording or consenting. A general platform comparison cannot supply private project terms.
Compare the payment unit with all your working time
A headline hourly rate and a fixed task payout describe different things. Establish which time is payable under the actual agreement, then include your preparation and corrections when comparing what the work would mean for you. Use one currency and avoid counting a fee twice.
| Assumption | Hourly example A | Fixed-task example B |
|---|---|---|
| Amount under these invented terms | $20 for 60 eligible minutes | $18 if the task is accepted |
| Other entered costs | No fee entered | $1 fixed fee |
| All working time | 60 eligible minutes + 30 unpaid preparation = 90 minutes | 45 doing/checking + 15 preparation + 15 corrections = 75 minutes |
| Net before tax | $20 | $18 − $1 = $17 |
| Effective hourly equivalent | $20 ÷ 1.5 hours ≈ $13.33/hour | $17 ÷ 1.25 hours = $13.60/hour |
B has the slightly higher effective equivalent under these assumptions, despite its smaller headline payout. That does not make it a better real offer: rejected work, different paid-time rules, extra corrections, other costs or unavailable work can change the result. The unpaid preparation in A is an invented condition, not a statement about any named platform.
To check B, open the task payout checker, select USD, and enter 18 payout, 45 task minutes, 15 preparation minutes, 15 correction minutes and a 1 fixed fee. The result is 17 before tax across 75 minutes, or 13.60 per hour. The selector changes currency display only; it does not convert amounts. Use the planning mode separately when estimating a freelance rate to charge.
Distinguish a real cost from a deposit-to-unlock scam
The FTC's task-scam alert describes fake earnings displayed in an app, followed by demands for a deposit to continue or withdraw them. Do not send money to unlock a supposed balance. The FTC's related task-scam analysis describes small initial payouts used to build trust. An early payment does not establish that a later request is safe.
Ask what any charge actually covers. A payment-transfer fee or a device you already use is different from paying a stranger to release supposed earnings; a stated cost still does not verify the offer. Check project requirements before buying equipment or training, and do not assume that spending money will produce assignments. Spirelight's contributor signup is free.
Reach the organisation independently through its official site. Review the recipient and purpose before supplying identity or payout information through a verified process. DataAnnotation documents identity verification in its FAQ, so a request for identification is not by itself proof of a scam. Keep account passwords and security codes confidential; use a code you requested only in the verified flow that issued it. For more checks, read how to assess an AI training job offer.
Check whether the work fits your device and skills
An app, a mobile signup or a phone-based identity check does not establish that the actual work can be completed on a phone. RWS's community FAQ, for example, lists a personal Windows or Mac computer, internet access and a current Chrome browser among its equipment requirements. Read the specific role's instructions rather than treating “remote” as “phone-only.”
For speech work, compare the recording brief with your device, location and permitted recording setup. For transcription or annotation, check the required software, writing conventions and review process. Avoid buying a microphone or laptop on the assumption that it will unlock better-paid work.
Use English transcription practice with audio and answers to try the rules shown there, and the transcription equipment checklist to consider your setup. Practice and equipment advice do not qualify you for a third party's assignment or guarantee an invitation.
Where Spirelight fits in a platform shortlist
Spirelight's contributor opportunities concern speech recording, transcription, transcript checking and audio annotation. A free profile provides information for potential matching; availability depends on your languages and the requirements of projects. Registration, practice and assessments do not guarantee work, an accepted submission or income.
The English signup form asks for legal names, email, a password, date of birth, gender and languages, with required fields marked. Review the separate email and WhatsApp offer preferences, which start selected, and the terms and privacy policy before agreeing. The initial form does not establish which later checks or project-specific terms might apply.
Use one truthful profile if this type of work fits. Read about Spirelight and its team and how its speech projects work before deciding. Questions about a specific invitation belong with its independently verified official project channel.
Keep your shortlist manageable
Choose the number of applications you can assess and maintain accurately. Note each platform's requirements, qualification time and unresolved questions; avoid opening duplicate accounts to pursue the same work. Applying to several providers can add administration without producing assignments, and a good record on one site does not establish transferable status on another.
Return to the actual project terms when an offer arrives. If you first need an overview of recording, transcription and annotation tasks, use the broader guide to getting paid to train AI as a freelancer. This comparison helps you ask informed questions; it does not certify a provider or predict your earnings.