The Credential Premium: What a JD, MD, PhD or CPA Is Actually Worth Per Hour
2026-08-27 · Expert Match AI team
Everybody in this market knows credentials pay more. Nobody has priced them.
The field averages on our salary report tell you what *categories* pay - legal $196/hr, healthcare $119, science $111 - but a category is not a credential. "Legal" includes a $40/hr annotation queue. "Healthcare" includes a medical coder. What a searcher actually wants to know is narrower and more useful: if I hold this specific qualification, what is the hourly rate attached to it?
So we went through all 313 live roles on the board, isolated the ones whose titles name a specific credential, and computed the midpoint average for each. Here is what the letters after your name are worth in AI training work, as of this morning.
The credential leaderboard
| Credential | Priced roles | Avg rate | Range | |---|---|---|---| | JD / attorney | 22 | $198/hr | $60–$1,000 | | MD / physician | 6 | $155/hr | $15–$450 | | PhD / professor | 3 | $229/hr | $100–$300 | | CPA / accountant | 6 | $77/hr | $40–$120 | | CFA | 0 | — | — |
Board-wide, the 209 roles with a disclosed hourly range average $92/hr. So a law degree is worth roughly 2.2x the market, and an accounting qualification is worth slightly less than the market.
Three things in that table are worth stopping on. A fourth - the PhD row - is worth treating with suspicion, and we will come back to why.
A JD out-earns an MD by $43/hr, which is backwards
In the ordinary economy, physicians out-earn lawyers, and it isn't close. Median physician compensation in the US runs well ahead of median attorney compensation, and the gap widens at the specialist end.
In AI training work, the ranking flips. Attorney-titled roles average $198/hr against $155/hr for physician-titled roles - and the attorney sample is nearly four times larger, so it is much the more reliable of the two numbers.
The reason is not that labs value lawyers more than doctors. It is that legal reasoning is the thing frontier models are currently worst at in a way that is *expensive*. A model that hallucinates a case citation creates liability the moment it reaches a client. Labs are buying error-detection in the domain where undetected errors cost the most, and the going rate reflects that urgency. micro1 alone lists six separate attorney specialisations - BigLaw, corporate, family law, immigration, litigation associate and M&A - each posted at $140–$400/hr. That is a company buying legal judgment by the practice area, not by the person.
Above all of them sits Surge's attorney tier at $500–$1,000/hr, the highest listed rate anywhere on the board.
Medicine has a high ceiling too - Surge's physician tier runs $250–450/hr and Mercor's Physician Talent Network $110–250 - but far fewer doors, and a much lower floor. Six priced physician-titled roles against twenty-two attorney ones, and the physician range starts at $15/hr for a radiologist listing whose floor we would treat as a placeholder rather than an offer.
The CPA discount is real, but it has narrowed
An accounting qualification averages $77/hr here - below the board average, and well under half what a law degree commands.
That is, however, a notably better number than when we first ran this analysis in mid-August, when accounting sat at $65/hr and looked like the worst-compensated major credential on the board. Two things moved it: the bottom of the range has firmed up, and micro1's Tax Advisor (CPA) role at $80–110/hr arrived to pull the middle up. The ceiling is still Alignerr's Finance & Accounting Task Author at $60–120, and the floor is micro1's Accountants and Auditors listing at $40–95.
The underlying story hasn't changed even though the number has. Accounting is highly credentialed, highly regulated work - and models are already fairly good at it, because it is rule-following applied to structured data, which is the shape of problem machine learning handles well. The scarce input in AI training is judgment where the rules run out. Accounting has fewer places where the rules run out than litigation does.
Zero roles ask for a CFA
We searched all 313 listings. Not one names the CFA charter.
That is worth saying plainly because the charter is routinely listed in AI-training round-ups as a qualifying credential, ours included - our own finance field guide told CFA charterholders to put the letters in the first line of their resume. That advice is still reasonable for automated screeners parsing a resume, but nobody should read it as evidence that CFA-gated roles exist on this board. They don't, today.
What *does* exist, and is new this week, is micro1's Investment & Finance Expert role at $245–280/hr - the highest-paying finance listing we have ever tracked, and one that names no credential at all. It asks for the experience, not the letters.
Where PhDs sit, and why we don't trust the number
The PhD row says $229/hr. Do not believe it.
Only three PhD-titled roles on the board currently disclose an hourly rate, and all three happen to be premium tiers: micro1's PhD & Academic Expert at $245–280, Surge's Professors & PhD Researchers at $150–300, and a Child Development Psychologist (PhD) role at $100–300. Meanwhile Turing lists at least six PhD-gated science roles - biology, chemistry, physics, advanced maths - that publish no rate at all and therefore cannot enter the average.
So the honest reading is that the doctorate average is unmeasurable right now, and $229 is what survivorship bias looks like in a small sample. The broader science field average of $111/hr is the number a PhD should plan against, with premium tiers as upside rather than expectation. Priced science work that doesn't require a doctorate sits considerably lower - research assistant roles run $30–50/hr. We covered the mechanics in the PhD field guide.
What the premium is actually paying for
Read the table again and a pattern shows up that has nothing to do with prestige.
The credentials that pay are the ones where a wrong answer is both expensive and hard to spot. Litigation, clinical medicine, frontier physics. The credentials that pay less are the ones where the work is rule-application - accounting, coding, translation - which is precisely where models have already closed most of the gap.
This has a practical consequence if you hold more than one qualification. Lead with whichever of them sits closest to genuine ambiguity. A physician who also holds an MBA should be applying as a physician. A lawyer who also codes should be applying as a lawyer, not as an engineer, despite software being the deepest category on the board at 54 priced roles - because software averages $92/hr and BigLaw litigation averages four hundred.
Methodology
Credential figures are the midpoint average of every live listing whose title names that credential, restricted to listings that disclose an hourly rate. Title-matching is deliberately strict: it undercounts roles that require a credential in the body text without naming it up front, so treat these as conservative floors rather than ceilings. Samples of three and six roles move on a single listing - the JD figure is the only one here with enough listings to be genuinely stable, which is why we have flagged the PhD row as unreliable rather than quietly printing it.
Every figure here - credential, field and board - is computed on the same basis against the live board on 2026-08-27: the 209 listings of 313 tracked that disclose an hourly range, using range midpoints. Listings that quote annual or per-project pay, or a ceiling with no floor, are excluded rather than converted. The salary report recomputes on a six-hour cycle and may show slightly different field averages depending on where in that cycle you read it.
These are listed rates from open postings, not accepted-offer data; read what "listed rate" actually means before mentally spending it.
Every role behind this table is on the jobs board. You can upload a resume on the homepage to see today's board ranked against your own credentials - 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.