State of AI Training Pay: July 2026

2026-07-31 · Expert Match AI team

This is the first edition of a monthly series: a data report on what AI training and expert-evaluation work actually pays, computed from live listings rather than surveys or anecdotes. As of this morning we track 317 live roles across 10 platforms, of which 223 disclose an hourly rate - those 223 are the basis for every average below (midpoint of each posted range). The numbers refresh every 6 hours on the live salary report; this post freezes the July 31 snapshot and reads the month's story out of it.

The July leaderboard, by field

| Field | Priced roles | Avg rate | Range | |---|---|---|---| | Legal | 15 | $138/hr | $20–$1,000 | | Healthcare | 8 | $136/hr | $20–$450 | | Software | 86 | $84/hr | $4–$300 | | Finance | 8 | $72/hr | $7–$180 | | Science | 19 | $70/hr | $7–$300 | | ML, Data & AI | 60 | $65/hr | $2–$300 | | Linguistics | 20 | $46/hr | $10–$400 | | Business | 7 | $43/hr | $7–$130 |

Legal keeps the crown it has held all month, but barely. The gap between legal and healthcare, which stood around $18/hr in mid-July, is now $2/hr.

The month's movers

Compared with our mid-July baseline (218 priced roles on July 14):

Healthcare, up ~$17/hr (≈$119 → $136). The story of the month. Premium clinician listings - Surge's physician tier at $250–450/hr, Mercor's specialty MD networks at $170–250/hr - kept arriving while lower-paid roles got absorbed off the board. We covered the mechanics in yesterday's healthcare deep-dive: fewer open seats, near-top rates, and the fastest absorption we track.

Software, up ~$8/hr (≈$76 → $84). The biggest category on the board (86 priced roles) also got richer, driven by senior-tier evaluation work: multiple $150–200/hr listings at Braintrust and Mindrift, and micro1's new AI trainer role listed at $100–180/hr. Senior engineers out-earn juniors more in this market than in salaried work - see the engineering field guide.

Finance, up ~$8/hr (≈$64 → $72). A small field (8 priced roles) so single listings move the average, but Mercor's Financial Analyst network at $60–180/hr is doing real work here.

Linguistics, down ~$19/hr (≈$65 → $46). The one big decline, and the composition explains it: micro1 shipped a wave of audio-recording roles this week (Dutch, German, Italian, Mandarin, Brazilian Portuguese) listed at $20–40/hr. That's new supply at the entry tier, not pay cuts to existing roles - the field's ceiling is still $400/hr for premium work like Surge's journalist tier. If anything the wave supports the bilingual-niche thesis: language demand keeps broadening.

Legal (+$1) and science (-$4) were effectively flat; ML/data held at $65/hr.

By platform: the 12x spread

| Platform | Priced roles | Avg rate | Range | |---|---|---|---| | Surge | 4 | $406/hr | $150–$1,000 | | Braintrust | 24 | $105/hr | $10–$260 | | Mercor | 18 | $102/hr | $60–$250 | | Stellar | 4 | $87/hr | $25–$200 | | Mindrift | 9 | $71/hr | $37–$200 | | micro1 | 144 | $65/hr | $2–$300 | | Alignerr | 10 | $64/hr | $20–$120 | | DataAnnotation | 5 | $52/hr | $20–$100 | | Outlier | 5 | $32/hr | $15–$50 |

The spread between the highest- and lowest-averaging platforms is roughly 12x, which remains the single most consequential fact in this market: where you apply matters more than what you know. micro1 is the volume engine - 144 of the 223 priced listings, nearly two-thirds of the board - while Surge lists only 4 roles and pays like no one else. One mid-month shift worth noting: micro1's average climbed from ~$57 to $65/hr as senior engineering roles joined its board, while Mercor's eased from ~$116 to $102/hr as its mix broadened. The playbook conclusion is unchanged: serious earners work 2–3 platforms at once, pairing a premium queue with a volume queue.

Superlatives, July 2026

• Biggest listed rate we saw: Surge AI - Attorneys & Senior Executives at $500–$1,000/hr, the only four-figure hourly on the board.
• Deepest field: software, 86 priced roles - more than legal, healthcare, finance, and business combined.
• Notable arrivals this week: micro1's multilingual audio-recording wave ($20–40/hr, remote worldwide), micro1's Go/Python/TypeScript engineering role ($100–130/hr), and Mercor's Writing Expert ($75–100/hr) and Mechanical Engineering Expert ($70–80/hr) networks.
• Steadiest number in the market: legal's field average, which has moved about a dollar in three weeks.

Methodology, briefly

Averages use the midpoint of each listing's disclosed hourly range, across every priced role live at the snapshot time; per-project and undisclosed-rate roles are tracked but excluded from averages. This is listing data, not accepted-offer data - read what "listed rate" actually means before mentally spending it. Mid-July comparisons come from our July 14 snapshot of the same pipeline; small fields (finance, business, healthcare) can swing on a handful of listings.

The August edition will publish at the end of next month. Until then the salary report recomputes every 6 hours, the jobs board carries every role behind these numbers, and you can upload a resume on the homepage to see today's board ranked against your background - free, no account, resume never stored.

Published by the Expert Match AI team. Pay figures are listed rates from live postings, not a survey of accepted offers. Some outbound application links on this site carry disclosed referral codes; rankings and recommendations are never influenced by them.