The comforting version of this market is that the generalists do code, maths, law and medicine, and creative work is an open lane. The comforting version is wrong. Mercor runs a dedicated creative vertical, is already buying the highest-value asset in the category, and is paying more for it than the specialist advertises.
Mercor is selling the moat, today, at the top of the band
Mercor's creative page headlines "Earn $60–$120/hr" (mercor.com/experts/creative). The headline understates the live postings.
| Role | Published rate | What it is |
|---|---|---|
| Senior Design Expert — Paid AI Design Research Study | $150–$250/hr | reasoning capture, recorded remotely |
| Professional Design Experts | $80–$180/hr | panel evaluation |
| Agency Brand Design Expert | $80–$150/hr | rubric authoring + scored critique |
| Media/journalism/communications Evaluator | $80–$120/hr | |
| Visual Quality Expert — Film, VFX & Animation | $60–$90/hr | |
| Character and Facial Animation Consultant | $60–$90/hr | |
| Design Expert | $70–$80/hr | |
| Graphic & UX/UI Designer | $60–$80/hr | |
| Graphic Designers (US/UK/CA, min 15 hrs/wk, now closed) | $60–$80/hr | "top AI company", ran through mid-2026 |
| Music & Lyrics Experts | $14–$78/hr | banded by language |
| Music Production Experts | $11–$60/hr | banded by language |
Sources: mercor.com/experts/creative; Senior Design Expert; Agency Brand Design Expert; Graphic Designers.
Read the two specs, not the rates.
The Senior Design Expert posting is the trajectory product. "The client wants to capture how expert designers think: how you approach a problem, what separates strong craft from weak" and when work is ready to ship. Four to five hours per task or interview, remote, structured and recorded, portfolio required (Mercor).
That is narrated expert design reasoning — the asset The trajectory moat identifies as the one thing in this category with a real cost floor, and the asset Contra's dossier identifies as Contra's only durable moat. It is being bought right now, at $150–250/hr, by an unnamed frontier client, through the largest vendor in the sector.
The Agency Brand Design Expert posting is the rubric product. The work is explicitly not producing design: "Design precise, task-specific grading criteria" and "Score AI-generated and human work samples against those criteria, with detailed written justifications." Requirements: 5+ years, "proficiency in typography, layout, color theory, and design critique", "exceptionally strong written communication", openness to "peer calibration", and agency pedigree — "(e.g. Pentagram, Wolff Olins, Landor, Collins, IDEO)" (Mercor).
Rubric design, written justifications, peer calibration, brand-agency credentialing. That is the specialist pitch, sold by a generalist, with the pedigree filter already in the job ad.
The lane is not empty. It is occupied by the biggest player in expert data, at the top of the price range, with the good version of the product — recorded reasoning and calibrated rubric authoring, not comparison volume.
Contra's "up to $100/hr" is mid-band
| Tier | Rate | Evidence |
|---|---|---|
| Senior design reasoning capture (Mercor) | $150–$250/hr | company posting |
| Agency-pedigree brand critique (Mercor) | $80–$150/hr | company posting |
| Contra Labs headline | up to $100/hr | contralabs.com/jobs |
| Contra × Lica TASTE study, actually paid | ~$90/hr flat project fee | arXiv 2605.20731 |
| Working graphic designer (Mercor) | $60–$80/hr | company posting |
| Market median, all AI training work | $65/hr | aitraining.jobs |
| Outlier "expert" tier | $35–$65/hr | third party [UNVERIFIED] |
| Surge contractor baseline | $18–$24/hr | Sacra |
"Up to $100/hr" reads as a premium against the $65/hr market median tracked across 1,697 open roles on 58 platforms (aitraining.jobs). It reads as fill-in work against a Pentagram-calibre designer's client rate, and as the middle of the band against Mercor. To out-recruit Mercor for the top decile the client needs $150+/hr, or the non-cash terms Contra uses — licence not assignment, byline, seven-day pay, no platform fee, right to decline — and realistically both. Will they say yes argues those terms are cheaper and more effective than a 30% rate premium, because for a large minority of creatives the objection was never the money.
The seam: Mercor can source designers but does not staff design
Mercor's own Ashby board carries zero creative roles — Infrastructure Engineer ($130–500K), Strategic Project Lead ($120–200K), Data Scientist, Software Engineer–Agents, Product Manager ($180–300K), Research Operations–Code, Data Engineer (Ashby posting API). The creative vertical is run by generalist project leads.
Mercor can find the designers. What it does not have in-house is anyone who can design the study — pick the axes, build the rubric, decide what counts as a refinement failure. That is the seam, and it is narrow: a generalist at Mercor's scale hires that capability the quarter it becomes worth having. Phase decomposition is the concrete version of what "designing the study" means.
Terms are worth knowing because they set the recruiting comparison: independent contractor, some roles W-2 through Cincinnatus LLC, weekly payment via Stripe or Wise, most projects 15–20 hrs/week, "Projects can be extended, shortened, or concluded early", and no H-1B or STEM OPT support (mercor.com/experts/creative).
Everyone else, honestly
Surge — $1.2B annualised revenue (2024), bootstrapped and profitable, 130 full-time staff and ~50,000 contractors serving roughly 12 labs; contractor pay "30-40 cents per working minute", i.e. $18–24/hr, the floor of this market (Sacra). Creative exposure is by modality, not domain: they annotate images and transcripts, and Hemingway-bench compares "5,000+ expert judge assessments" on writing dimensions including creativity and humour.
I found no Surge design, art, illustration, photography, motion or 3D vertical. Surge publishes very little, so this is absence of evidence rather than evidence of absence. Do not brief it as "Surge is not in creative"; brief it as "Surge does not say".
Scale — Meta paid $14.3B (Forbes, Jun 2026). Across roughly 200 postings on its Greenhouse board there is exactly one design role, a Forward Deployed Product Designer (Greenhouse API). Scale is not building a creative vertical. Outlier contributor rates are reported third-hand at $15–65/hr [UNVERIFIED].
Handshake — 95 active listings analysed in August 2026 average $108/hr, range $22–$500/hr; Content Creation is 2 listings at an $85/hr average, and there are no dedicated video, graphic design or media production roles (AI Gig Jobs) [UNVERIFIED] — third-party listing analysis with a stated sample. The $300–500/hr tiers go to physicians, STEM researchers and radiologists.
That last fact is the most useful comparative in this page: the platform paying the highest average rates in AI training pays its top rates to doctors and has no design vertical at all. Whatever a designer's judgement is worth to a lab, nobody has yet priced it like a radiologist's. Eleven labour markets, not one is where that ceiling gets argued.
micro1 — expert categories are "engineering, finance, healthcare, legal, economics, linguistics, and cinematography", with an example role of "Emmy Award-Winning Cinematographer"; vetting is an AI interviewer called Zara; "Certification does not guarantee placement" (micro1.ai/experts). No creative pay rate is published on micro1's own site — the $30–65/hr figure circulating in competitive decks comes from the aggregator aitraining.jobs, not from the company.
Turing — the commodity tier. A live Image/Video Annotator posting asks for annotation of "subjects, settings, actions, and emotions" and labelling of "visual cues like mood and lighting", minimum 20 hrs/week, contractor, no benefits, no published rate, and "prior image/video annotation experience preferred but not required" (Turing).
"Preferred but not required" is the line between the tiers. Turing's visual work is annotation. Mercor's is critique. The client's business only exists on Mercor's side of that line, which means the relevant competitor set is one company, not five.
What this does to pricing
Two things, and they pull in opposite directions.
Supply-side, the clearing price for creative judgement is now publicly legible in a way it was not six months ago, and it is set by Mercor rather than by the specialists. Any recruiting pitch that leads with a rate is arguing on a number a competitor can beat by 50% the same afternoon, which is why the durable pitch has to be terms — see The clause that expires your corpus in year ten and Where the good ones actually are.
Demand-side, nothing is legible at all.
Not one company in this field — generalist or specialist, Mercor, Surge, Scale, Handshake, micro1, Turing, Contra Labs or Taste Labs — publishes a buy-side price for a creative-preference dataset or an evaluation engagement. Every rate above is what the worker is paid. The margin between the two is the whole business and it is invisible in the public record; the honest working assumption is a 1.5–2.5× markup, which is an industry norm and not a measurement. GMV is not revenue applies to every revenue figure quoted in this section.