RESEARCHYou Can't Buy Taste. We Tested It.

Handshake AI vs Mercor: one of them may be supplying the other

Most comparisons in this category put two vendors side by side as substitutes. This one needs a different frame, because of something a Handshake executive said on the record in January 2026: Mercor and Scale AI often use Handshake's platform to source the human experts they resell. If that is accurate, buying from Mercor sometimes means buying Handshake's supply with a markup on top, and the choice between them is partly a choice about how many layers you want in the chain. This comparison works through what each actually operates, where the evidence is thin on both sides, and which one fits which kind of program.

Key takeaways

  • Handshake runs a career network of 20 million students and professionals and recruits AI work from inside it. Mercor recruits, screens by AI interview, and matches experts to projects for a markup.
  • Cleanlab's chief executive, now leading research at Handshake, has said publicly that Mercor and Scale often source experts through Handshake's platform. Treat that as a competitor's claim, but it is on the record and unrebutted.
  • Handshake's fellowship is United States only. Mercor publishes 107 supported payout countries. For multilingual or regional work this is decisive.
  • Both report revenue as gross marketplace volume. Neither publishes net, and the gap is large enough that the headline numbers mislead.
  • Mercor's compliance posture is thin but real and covers Mercor. Handshake's trust page describes its education platform, not its AI data business.
  • Neither publishes a customer case study with attributable metrics, and neither has buyer-side review coverage anywhere.

About Handshake AI

Handshake was founded in 2014 by three Michigan Tech computer science students and spent a decade as a campus recruiting network, reaching a stated 20 million students and professionals across more than 1,600 schools. Handshake AI launched in January 2025 as a second business built on that network, selling reinforcement learning environments, evaluations, and human data to frontier labs.

The pivot has been fast and not painless. In October 2025 the company cut roughly 100 roles, about 15 percent of staff, from the recruiting side, with chief executive Garrett Lord describing the move as a refounding. The AI unit grew from 15 people to about 150 in the same period.

Its supply is the career network itself. Handshake states 188,000 or more fellows, a community of 750,000 or more students, professionals, and experts, and more than 500 million dollars in payouts, all self-reported. Lord has described the underlying pool as 500,000 PhDs and three million advanced degree holders. The fellowship publishes rates of up to 100 dollars per hour.

One constraint matters more than any of those numbers. The fellowship requires United States work authorization, so the delivery workforce is domestic only.

In January 2026 Handshake acquired Cleanlab, a nine-person team of MIT computer science PhDs that had raised 30 million dollars, in what was reported as an acqui-hire. Cleanlab's open source package carries more than 11,700 GitHub stars, now under Apache 2.0, and its Hugging Face organization holds 28 datasets. Handshake acquired the AI skilling company Uplimit in June 2026.

About Mercor

Founded in January 2023 by Brendan Foody, Adarsh Hiremath, and Surya Midha, based in San Francisco. It raised a 100 million dollar Series B at 2 billion in February 2025 and a 350 million dollar Series C led by Felicis at 10 billion in October 2025. A further round at roughly 20 billion has been reported in talks since July 2026 and has not closed.

The model is an expert marketplace. Candidates complete an AI-run interview of about twenty minutes, with up to three retakes, and qualified experts are matched to client projects with Mercor billing a marked-up hourly rate. Mercor states 30,000 or more active experts and an average contracted rate of 107 dollars per hour, with published listings from around 60 dollars per hour for language tutors to 450 for oncologists. Its own documentation lists 107 supported payout countries across Africa, the Americas, Asia, Europe, and Oceania.

Mercor sells beyond staffing as well: a data monetization line that licenses enterprises' own operational data to AI labs after extraction and de-identification, off-the-shelf data, and a robotics line that licenses wearable-camera footage captured at partner businesses. It publishes the APEX benchmark family alongside partner benchmarks built with Harvey, Ramp, Cognition, and Sierra. It acquired Sepal AI in February 2026 and Deeptune in July 2026.

What buyers report

Neither company has usable buyer-side review coverage, and in Handshake's case there is none at all.

Handshake AI has no G2, Gartner Peer Insights, Capterra, or TrustRadius profile. The Handshake listings that do exist cover the career network: 4.5 out of 5 on G2 across roughly 78 reviews for Handshake for Employers, and a Gartner listing for Handshake Premium showing 3.9 from five ratings. Those measure a campus recruiting product and tell you nothing about frontier data delivery.

Mercor holds 4.4 out of 5 on G2 across 114 reviews, but its profile sits under applicant tracking, AI recruiting, and job search categories, so the reviewers are candidates and contractors rather than enterprises buying data. It has no Gartner Peer Insights listing.

For both, the public ratings measure the labor-supply side. Plan on a paid pilot.

The comparison

Handshake AIMercor
OriginCareer network since 2014; AI business launched January 2025Purpose-built marketplace since January 2023
Supply188,000-plus fellows drawn from a 20 million-person network30,000-plus active experts, recruited directly
VettingProject-specific screening; no published methodologyAI interview, roughly 20 minutes, up to three retakes
GeographyUnited States only107 supported payout countries
Published ratesUp to 100 dollars per hour107 dollars per hour average; listings 60 to 450
ModalitiesText and code; limited visionText and code; plus wearable-camera video via partner businesses
ComplianceSOC 2 Type II claimed, TX-RAMP, UK Cyber Essentials, scoped to the education platformSOC 2 Type 2, via a trust center covering Mercor
Funding200 million dollar Series F, January 2022, around 3.5 billion valuationRoughly 484 million raised; 10 billion priced October 2025
Open researchSeven public benchmark repositories plus Cleanlab's 11,700-star packageAPEX benchmark family
IP indemnificationNot published; no public terms of serviceNot published; public terms indemnify Mercor only

The supply-chain question

In TechCrunch's coverage of the Cleanlab acquisition in January 2026, Cleanlab chief executive Curtis Northcutt, who now leads AI research and strategy at Handshake, said that Mercor and Scale AI often use Handshake's platform to source human experts such as doctors, lawyers, and scientists. His framing was direct: if you are going to pick one, pick the source rather than the middleman.

Weigh that for what it is. It is a competitor's characterization, made by someone who had just joined Handshake, and Mercor has not publicly rebutted it. It is also plausible on the structure: Handshake owns a credential-verified network of 20 million people that took a decade to build, and Mercor's business is finding qualified people quickly. Nothing stops a marketplace from recruiting through another network.

What a buyer should do with it is ask. If you are evaluating Mercor for a program that needs credentialed specialists, ask where its supply originates and what share is sourced through third parties. If the answer involves Handshake, you are paying two margins for the same people, and the question becomes whether Mercor's matching, project management, and accountability are worth the second one. For some programs they clearly are. For a team with its own task design and quality process, they may not be.

Geography is the cleanest differentiator

Handshake's fellowship requires United States work authorization. That is not a soft preference; it is the stated eligibility rule, and it has shown up in contractor disputes where work performed outside the US was cited as grounds for non-payment.

For a buyer this cuts two ways and both are real. If your data has residency constraints, if you need US persons for regulatory or contractual reasons, or if you want a workforce under a single well-understood labor regime, Handshake's constraint is an advantage and most competitors cannot match it. If you need Hindi, Arabic, Portuguese, or Bahasa evaluation, or region-specific professional judgment, Handshake structurally cannot serve you and Mercor's 107-country footprint can.

Reading the revenue figures

Both companies are described with numbers that are not what they appear to be.

Handshake's company-wide revenue was estimated at roughly 1.1 billion dollars annualized in April 2026, growing 349 percent year over year, with the AI segment at nearly 1 billion gross. Those are gross marketplace figures. The same analysis puts company-wide net revenue at roughly 450 million and the AI segment's net at roughly 300 million, on gross margins near 30 percent, against roughly 80 percent for the legacy job board.

Mercor's widely cited 2 billion dollars is gross annualized revenue on the same basis, with contractor share estimated at 60 to 70 percent. Mercor does not publish net, and outside estimates span 180 million to 800 million.

Both sets of figures are third-party estimates rather than company disclosure. Neither should be used as a proxy for stability. The funding record is documented and the customer evidence is not, so weight the former.

Published research, where Handshake is ahead

This is the dimension where Handshake has something Mercor does not match.

Its research organization publishes seven public benchmark repositories under permissive licenses, including an agent-as-a-judge grading framework and an investment-banking workflow benchmark, and two substantive papers: one on verifier design and one finding that more than 80 percent of coding agent rollouts reason about an imagined grader rather than the user. The Cleanlab acquisition added an open source package with more than 11,700 stars and 28 public datasets.

Mercor publishes the APEX benchmark family and partner benchmarks with Harvey, Ramp, Cognition, and Sierra, which is real work, but the open footprint is smaller and there is no public dataset organization.

Published benchmarks matter to a buyer because you can inspect the task design and judge whether the evaluation rigor on display resembles what you would get on your own project.

Compliance, where neither is strong and one is confusing

Mercor operates a trust center listing a SOC 2 Type 2 report, penetration testing, business continuity plans, cyber insurance, and subprocessors. ISO 27001 appears in its job postings as a framework being stood up rather than achieved. HIPAA and GDPR attestations are absent. It is a thin posture, but it is documented and it covers Mercor.

Handshake's trust page claims SOC 2 Type II along with TX-RAMP certification, UK Cyber Essentials, and PCI DSS self-assessment at SAQ-D level. The problem is scope: everything described covers the education and career platform, with students, employers, and universities as the subjects, and nothing states that any certification extends to Handshake AI or the data business. ISO 27001 appears only as something its cloud host holds. The page is also internally inconsistent, claiming ten certifications while listing seven. Handshake AI publishes no terms of service, and the expected URL returns an error.

Neither publishes IP indemnification or data-ownership terms. Ask for both in writing before scoping, and ask Handshake specifically whether its attestations cover the AI business.

Which to choose

Choose Handshake when the work requires United States persons, when credential verification against an institutional record matters more than speed, or when you value the research signal: this is a team publishing benchmarks and verifier research rather than only selling hours. Ask what its SOC 2 actually covers.

Choose Mercor when you need coverage outside the United States, when speed to staff is the binding constraint, or when you want the adjacent lines Handshake does not offer, including data monetization and wearable capture. Ask where its expert supply originates.

Why HumanSignal is the better choice when the data has to be produced

Both companies here answer the same question: how to find qualified people quickly. Handshake answers it with a decade-old credential network, Mercor with recruiting and AI screening, and if Northcutt is right the two answers are partly the same answer sold twice. Neither addresses what happens when the examples your model needs have never been recorded by anyone, which is the binding constraint in robotics, embodied AI, and any domain where the knowledge lives in physical practice rather than in text.

HumanSignal Services is built for that case. We write the collection protocol, recruit against the characteristics it requires, run the capture in our studios, on site at your facility, or in the field with synchronized motion, vision, audio, and sensor streams, and deliver a dataset verified against the spec. The accountability for the result sits with us, not with a sourcing layer, which is the difference between commissioning a dataset and renting the people who might produce one.

The quality mechanism is inspectable. Review workflows run in Label Studio Enterprise, the commercial edition of the open source labeling tool we build, so inter-annotator agreement, reviewer decisions, and rework rates are reportable to you while the work is running. We publish contributor pay ranges, which makes our sourcing claims checkable rather than rhetorical. Our customer agreement gives you exclusive ownership of customer data and of insights produced specifically for you, which is more than either company here publishes. We hold SOC 2 Type 2 and HIPAA compliance, documented on our security page, and the scope is the company rather than an adjacent product line.

The capture operation came in through the acquisition of Erud AI in November 2025. If your constraint is producing data rather than sourcing people, physical AI data services sets out how a program gets scoped.

Does Mercor really source experts through Handshake

It has been stated publicly by Cleanlab's chief executive, now Handshake's head of AI research, and reported by TechCrunch in January 2026. Mercor has not publicly rebutted it. Treat it as an unrebutted competitor claim rather than an established fact, and ask Mercor directly where its supply originates if the answer would change your decision.

Why does Handshake only work with US-based contractors

Handshake has not explained the restriction. The fellowship states that eligibility requires United States work authorization. The likely drivers are the domestic campus network the business is built on and the payment and tax infrastructure that comes with it. The practical effect is that non-US-language and region-specific work has to go elsewhere.

Which has the larger expert network

Handshake, by its own figures, though the two count differently. Handshake states 188,000-plus fellows drawn from a 750,000-person community; Mercor states 30,000-plus active experts. Handshake's figure counts people enrolled in a fellowship rather than people actively delivering, so the comparison is not like for like.

Can either collect physical or robotics data

Mercor can, in a limited way, through its robotics line licensing wearable-camera footage from partner businesses. Handshake publishes no physical collection capability at all. Neither designs capture protocols, so for instrumented or teleoperated collection you would be looking at collection specialists.

Related Content