The $31/hr Average: Why It Misleads AI Training Workers

The figure cited most frequently in AI training pay coverage — approximately $31/hr as a national average in the US — appears in ZipRecruiter, Indeed, and Glassdoor data. Here's why it's the least useful number to anchor expectations on, and what to use instead.

What the $31/hr Average Is Actually Measuring

This blended mean covers all job listings in the "AI trainer" category — including full-time employed AI trainers at tech companies, part-time freelance contributors on annotation platforms, specialist domain experts billing at $150+/hr, and data labelers earning $15/hr on microtask platforms. These are not comparable roles, and blending them produces a number that accurately describes none of them.

The Distribution Is What Matters

A more useful picture is the distribution across our specific platform reviews: generalist work on DataAnnotation.tech starts at $16-20/hr; mid-level evaluation on Outlier AI runs $25-40/hr; specialist domain work on SME Careers reaches $50-130/hr; and Mercor's platform-wide contractor average, per Contrary Research's February 2026 data, sits at $95/hr. The $31/hr figure sits at roughly the midpoint of this range — meaning it overstates what generalists will earn and understates what specialists can reach.

The $31/hr average is technically accurate as a statistical measure and practically useless as a personal benchmark. It is the average of wildly different things presented as a single number.

More Useful Benchmarks by Tier

• Generalist, no professional credentials: $16-25/hr to start
• Writing, research, or academic background: $25-40/hr
• Verified professional credentials (law, medicine, engineering): $50-130/hr
• Senior coding experience: $60-200+/hr

These ranges, covered in detail in our main pay breakdown, are platform-verified rather than blended averages across incomparable categories.

The Bottom Line

Ignore the $31/hr headline figure when planning your income expectations. Use platform-specific, tier-specific data as your reference — your actual rate depends on your background, the platforms you are accepted to, and the task types you qualify for.

The Three Numbers Worth Knowing Instead

Rather than the blended $31/hr, here are the three specific figures that are actually useful for personal planning:

These are not estimates — they are platform-verified or Glassdoor-sourced figures from 2026. The $31/hr national average sits between the generalist floor and the mid-level tier. It's mathematically accurate as a blend and practically useless as a planning benchmark.

Why the Spread Keeps Growing

The gap between the generalist floor and the specialist average is widening, not narrowing — consistent with the xAI decision we cover in our xAI layoffs piece and the broader bifurcation we describe in our pay gap explainer. RLAIF absorbing the simplest evaluation tasks puts downward pressure on the generalist floor; growing demand for expert evaluation in specialist domains drives the ceiling higher. The $31/hr "average" is therefore not stable — the distribution around it is actively widening.

What Determines Your Actual Earnings

Three variables determine AI training income more than any other: credentials (which sets your rate ceiling), hours (which sets your volume), and platform selection (which determines whether your credentials are correctly valued). A specialist on the wrong platform earns generalist rates. The same specialist on Mercor or SME Careers earns their market rate. Use our Platform Picker to ensure your background is matched to the platform that will value it correctly.

Realistic Monthly Income Milestones

Month 1: €0-300 (approvals, learning curve, first tasks). Month 2: €200-800 (established on 1-2 platforms, speed improving). Month 3: €400-1,500 (quality scores built, better task matching). Month 6: €600-3,000+ for consistent 15-25 hrs/week (depending on credentials). Contractors earning above €2,000/month consistently share one characteristic: they applied to multiple platforms in their first two weeks rather than sequentially waiting for each approval before applying to the next.