The fastest-growing paid-task category: labeling, evaluation, and expert-domain work that trains AI models — verified rates and qualification requirements for each platform.
Editor's pick · refreshed November 2026
Hand-picked on pay, transparency, and reliability — never paid placements.
Kuro pays for passive data sharing — billed at an hourly or project rate, paid via Tremendous and Gift card or voucher. $1 per accepted hour of screen recording, with acceptance criteria unpublished
Per hourly or project rate. $10-$250/hr by expertise tier; each project posts its rate up front
AI-powered talent network that connects domain experts (software engineers, finance, healthcare, legal, etc.) with paid AI training platforms.
110 platforms listed, including 2 that cover ai training & data labeling as a secondary category.
Lightweight Windows and macOS desktop app that pays you to let it record your screen, computer inputs and metadata while you use the…
TIERED since mid-October 2026: Basic $0.10 ('New participants start on Basic'), Apprentice $0.25, Advanced $0.50, Pro $1.00 USD per eligible active hour
at an hourly or project rate · min payout None - 'There is no minimum withdrawal amount and no payout fee'; payments
Human-data brand of CipherSense AI Technologies Limited, a Lagos-based AI company (NDPC data-controller registration NDPC/DCP/13640)…
Company-official (AI Talents Program): Data Annotator $5-$15/hr typical; QA Reviewer $8-$20/hr; LLM Evaluation Specialist $12-$30/hr
per task · min payout Not published - no threshold, rail, schedule or currency anywhere on the site
AI data and language-services arm of Summa Linguae Technologies, a Krakow-based group (founded around 2009) that acquired the original…
USD 10/hr for approved voice recording and USD 10-12/hr for approved transcription (Kabyle and Hmong projects)
per task · min payout Not published anywhere. Study incentives are paid by PayPal within 10 days of…
Contributor arm of Deccan.ai, Inc., a Delaware-incorporated AI training-data company (Bay Area HQ, Hyderabad operational backbone…
No official rate table - site says only 'Payments vary by project: some are task-based, others hourly, and a few use a combined model. All payment details
per task · min payout Not published - no threshold figure exists anywhere public (homepage
Dip Audio (dip.audio, San Francisco) pays contributors to build speech-AI datasets - 'voice assistants, transcription, and accessibility'…
'Up to $30 per recorded hour' across all 13 conversation-partner roles on the jobs board (identical compensation block with performance incentives)
per audio hour · min payout No minimum - 'Payouts go out every Monday with no minimum threshold and no…
Earn With Aloe pays people to have recorded conversations with matched strangers so that speech and avatar models can be trained on natural…
GBP 20 per hour online, about GBP 30 per 90-minute recorded conversation; the London studio pays GBP 50 per hour the same day
at an hourly or project rate · min payout GBP 10, paid by Stripe straight to your bank account in local currency
Enterprise-AI agency whose 'data practice' pays people to record their own real work as AI training data, plus a separate expert annotation…
Capture: $10-$30 per accepted hour by occupation (cleaning/kitchen/farm/retail $12; warehouse/factory $15; trades, mechanics, barbers $20; vet/lab from $25
at an hourly or project rate · min payout Not published - approved hours accrue to a balance paid out through your own…
French contributor platform that pays people to film egocentric (first-person) video of everyday tasks for robotics AI training - cooking…
Not published - each mission page shows its own per-hour rate before you join, and rates depend on where you are based
per task · min payout A withdrawal minimum exists but its amount is shown only on the in-app…
Enterprise-AI-data vendor marketing RLHF, annotation and red-team services while recruiting remote 'AI Tutors' (independent contractors)…
Portal bands (checked 2026-11-05): AI Output Validation Expert $10-$15/hr; Training Data QA Analyst $10-$20/hr; LLM Performance Specialist $15-$30/hr
at an hourly or project rate · min payout Not published - no payment method, schedule, threshold
Voice calling app that pays users for their conversations, which (with consent) are used as AI training data by vetted partner companies.
No per-minute rate is published; earnings come from the revenue split on purchased calls — 'Typically it's 50/50
per task · min payout $5 — earnings credit to the Fluffle wallet once a session is approved and…
Glint is a French two-sided data marketplace: 'We pay humans to record the data frontier labs train on' - AI labs, robotics companies and…
No live rate card - page examples are marked illustrative (photo set 34 EUR, POV video 18.75 EUR, annotation 27.20 EUR, audio session 15.00 EUR)
per task · min payout 25 EUR AND 100 validated files (CGU, 2026-07-03)
Globik AI operates an enterprise data solutions business alongside a contributor network ('Globik Studio') for AI data annotation…
No public rates; work is project-based expert engagement (e.g., radiology reader studies, linguistics quality evaluation) with pay set per role.
at an hourly or project rate · min payout No minimum payout threshold or payout schedule is published on the public site
Gray Swan's contributor arm is the Gray Swan Arena at app.grayswan.ai - a crowdsourced AI red-teaming network run by Gray Swan Security…
Contest/pool model, not hourly: official 'prize pools from $20,000 to $170,000+'
per task · min payout $100 for pool-based prizes - below that
Handshake AI runs AI Fellowship and AI Trainer roles on top of the Handshake career network used by US universities.
Handshake's own site states $40-$125/hr; live October 2026 postings range from $40/hr (AI Evaluation Specialist) to $125/hr (Game Developer/Designer
at an hourly or project rate · min payout No minimum payout threshold is published
Harbor (harborml.com) is a 'Private AI Platform' combining a chat product (Harbor-1; Free/Pro $15/mo/Super $59/mo), an agent search API…
Not published - 'The tier sets the approximate payout' (Bronze/Silver/Gold) with no dollar figures anywhere; the homepage self-reports '$18,420
per task · min payout Not published - payouts land as USDC or Harbor API Credits
MacOS menu-bar application that records your screen and interactions while you work - 'and nothing leaves your Mac raw' - anonymizes the…
Not published - session 'Rewards' are sized by how productive the captured work is
monthly for passive device data · min payout The FAQ claims you 'can cash out any time once you cross $10'
Contributor platform and mobile data lab that collects real-world human experiences for AI and robotics training.
Advertised $7–$12 per recorded hour across 14 work sectors on the contributor site (e.g. $12/hr plumbing, HVAC & electrical; $9/hr auto services and restaurants
per task · min payout No minimum withdrawal amount is stated on any official page
Paid AI training contributor network run by HumanSignal, the maker of the open-source Label Studio.
Official example rates (US residents): $15-$35/hr data collection contributor, $22-$45/hr AI response evaluator
at an hourly or project rate · min payout No minimum payout threshold is published
Human-data marketplace ('The human data marketplace.
Capture: $12-$30 per accepted hour by occupation (cleaning/kitchen/farm/retail $12; warehouse $15; trades/mechanics/beauty $20; office/design $25
at an hourly or project rate · min payout Not published - Stripe payouts for accepted/recorded hours
Global AI data solutions company that hires data annotation, content moderation, and AI training professionals.
Advertised Scholar listings have shown roughly $7-9 per hour (e.g., $7/hr for a Spanish language specialist in Mexico, $9/hr for Vietnamese voice-testing work)
at an hourly or project rate · min payout No minimum payout threshold is stated
Julius Expert Earners is a paid contributor program run directly by Julius AI Inc.
“Every task pays a $100 baseline, with reward milestones on top”; company-quoted average ~$200/hour; tasks take 15 minutes to a couple of hours
per task · min payout Not published - experts are independent contractors 'paid by direct deposit'
Jwuma is the contributor platform of Corpshore AI, the data-operations arm of Toronto-headquartered BPO group Corpshore Solutions…
Field capture is illustrated at $5.00/task (8 tasks/day x 5 days = about US$200/week, 'an illustration, not a guarantee')
per task · min payout No numbers published - each rail has its own minimum payout threshold
Kled bills itself as 'The Leading Data Marketplace': contributors upload their own photos and videos and complete paid tasks, and accepted…
Dynamic - formally 'no flat per-file rate' (value set by uniqueness, quality and buyer demand; High/Medium/Low tier labels per task in-app).
per task · min payout $25.00 and ~100 accepted items; one-time government-ID + selfie KYC before…
LILT (San Francisco, founded 2015 by Spence Green and John DeNero; $92.5M raised) recruits paid freelance 'AI Data Experts' through…
Rarely published: 4 of 447 live Ashby postings state pay ($20/hr AI Training Contributor, $15-18/hr, $12-15/hr gaming coordinator, $10-15/hr)
per task · min payout Not published - compensation is issued within 15 days of completed work through…
LingoTribe was a voice-data platform by Mundo AI Inc.
No per-task rates are published; the official FAQ requires recorded conversations of at least 10 minutes to qualify for payment
per task · min payout No minimum payout threshold is published
Lionbridge's AI contributor platform hires freelance AI trainers, evaluators, linguists, and data annotators globally.
Advertised project rates are hourly — e.g., $7 USD/hour for a remote Hindi (India) audio recording project; rates vary by project, language and country.
at an hourly or project rate · min payout No minimum payout threshold is published on the official site
Aurora AI Studio is the contributor-facing side of Lionbridge's return to AI training data: a community claimed at '500,000+ users' across…
Not published - official language is only 'Earn competitive rates' (join-our-community) and 'Task-based compensation
per task · min payout Not published - no threshold figure on any Lionbridge/Aurora page or app…
Multi-modal AI data marketplace paying individual contributors for voice, video, and photo collection across many low-resource languages.
Advertised listings: about $12/hr ($0.20/min) for American English conversational audio and $7.50/hr ($0.13/min) for Uyghur, Tibetan, Hmong, Jamaican Creole
at an hourly or project rate · min payout No minimum withdrawal threshold is published
Clickworker takes beginners at microtask rates; Outlier pays expert-tier rates but gates entry with degree checks.
Read full comparison →AI training work pays people to improve machine-learning models: rating model responses, labeling images, writing and evaluating prompts, transcribing audio, and reviewing code. Demand has exploded, and so has the spread between the best and worst platforms — generalist microtasks pay a few dollars an hour while expert-domain projects (law, medicine, advanced coding) pay $15–40+/hr, sometimes on the same platform.
Every AI-training platform we list goes through the same verification discipline as every other category: pay models, qualification gates, identity requirements, payout methods, and country availability, checked against official contributor pages and terms. Where a platform hides its rates, we say so.
Before you apply, understand the two-tier structure. Generalist tiers accept most applicants and pay per task; expert tiers gate entry behind degree verification and domain tests but pay multiples more. The application-process row on each platform page tells you which you are walking into.
Expert tiers usually require at least an associate degree, with bachelor or higher for specialist domains. Generalist microtask tiers require no degree. Each platform page states the verified requirement.
Generalist work typically pays a few dollars per task; expert-domain projects pay $15–40+/hr. Across the 48 platforms we track, median expert-tier rates land in the $15–25/hr band.
Annotation labels raw data (images, text, audio); evaluation scores and ranks existing AI outputs. Evaluation work — especially prompt evaluation — is currently the most widely available expert-tier task.
Most expert-tier platforms verify government ID (often via third-party services like Persona or Onfido) plus tax information. Generalist platforms may only require email verification. Identity requirements are listed per platform.
Before you go
246 of 420 verified platforms pay via PayPal — this month’s stat, computed from our own tracked set. The full list, dead-platform warnings and the next stat land in your inbox at least three times a week.
No spam, no list sharing — unsubscribe anytime.
