Litmus

The compensation index for AI trainers.
We've collected what 449+ trainers actually get paid.

Compare Your Rate
Country filters drop in 536 submissions Get notified Submit You'll be first to know when country filters go live
All Micro1 Handshake Alignerr DataAnnotation Outlier Mercor
Engineering Math Science Finance Legal Medical Writing Generalist

Compare Your Rate

Have your platform open — we need the exact rate.

Why
Platforms profit from pay secrecy
3.6x pay gaps for identical work
Rate cuts happen without warning
This is the index they'll never publish
Your expertise deserves a price tag
Platform
Platform
Outlier
Alignerr
Handshake
Mercor
Micro1
DataAnnotation
Other
Project Name
Task Type
Task Type
Evaluation
Review
Annotation
Red Teaming
Other
Task Domain
Domain
Coding
Math
Science
Writing
Legal
Medical
Finance
Generalist
Other
Your Country
Country
Effective Hourly Rate
USD
Started at This Rate
Select Month
Thank you — you're now part of the index.

Evaluate

Prove what you can actually do.

Assessments
Engineering
16 questions · ~20 min
Math
Vote
Science
Vote
Finance
Vote
Legal
Medical
Vote
Writing
Vote
Generalist
Vote

Job Board

Real jobs. Real pay. No guessing what a listing actually pays or whether it's worth your time.

Min Max
Get the monthly rate report Never linked to your submission Send
Joining
Confirmed
First report drops next month

Feedback

Help us build something worth using.

Message

What's Next

Quality should be rewarded.
Something's coming for evaluators who agree.

Email

The Scouting Report

Five stories. Every week. No fluff.

Email
03-23-26
The Scouting Report #002
03-16-26
The Scouting Report #001

The Scouting Report #002

Email
Compare your rate

Pitch

The quality layer for human intelligence in AI training.

Why
AI labs spend billions on human evaluators but have no way to verify quality across vendors. Litmus is building the independent credentialing standard — the bar exam for AI evaluators.
Problem
A PhD mathematician and a mediocre one both show up through the same platform at similar rates. Labs are buying blind. Workers have no way to prove their value. Platforms capture a heavy margin as intermediaries.
Insight
Research shows high-quality annotations dramatically outperform volume, yet no portable quality credential exists anywhere. Every platform grades internally. None of those grades travel.
How it works
Phase 1 · Rate Index
Crowdsourced compensation data across platforms, domains, and geographies. Builds community trust and brand.
Phase 2 · The Combine
Paid competitive assessments with embedded bias diagnostics. Top 30% earn certification across competitive tiers. Labs pay for access to verified talent.
Phase 3 · Academy
Cohort-based training program for top performers. Direct pipeline to lab placements. Workers get paid, not charged.
Traction
650+
Data Points
5
Platforms Indexed
Revenue Model
12
Talent Intelligence
Labs pay for access to certified evaluator profiles
$10K per year
1234
Placement Fees
Direct matching of certified evaluators to lab projects.
$2–4K per head
12
Market Intelligence
Compensation benchmarks sold as annual subscriptions.
$20K per year
123
Sponsored Combines
Labs sponsor for branding, early access, and placement
$10K per cohort
1
Domain Expansion
Labs commission new Combine domains in specialties
$25K+ per domain
123
Quality API
Labs integrate quality score directly into their pipeline.
$5 per score lookup // $50K+ for enterprise integration
The Ask
$75K · $1M post-money cap · 7.5% dilution
Pre-seed SAFE to prove the certification model, validate lab-side demand, and build the dataset labs can't get anywhere else.
Capital goes directly into participant payments, grading operations, and domain question pool development. No salaries, no overhead — every dollar funds the proof.
This round proves three things:
  • Workers show up and compete for certification
  • The tiers produce meaningful, measurable quality differentiation
  • At least one lab pays for access to the certified pool
That proof unlocks a seed round to scale cohorts, expand domains, and build the quality API that makes Litmus infrastructure.
What validation looks like
300+
Certified evaluators across multiple domains
6
Monthly Combine cohorts with published results
1+
Lab LOI or pilot validating demand-side revenue