Handshake AI Review 2026: Is Handshake AI Legit? Yes. Now Beat the Timer.

Handshake AI moves more money to contractors than any platform we have reviewed, and it has the worst review pages of any of them. Both are true for the same reason.

The company that runs it, Handshake, was a college job board with 20 million users. In January 2025 it started selling those users' expertise to AI labs. Sixteen months later the AI unit was at roughly $1 billion in gross annualized revenue, eight frontier labs were buying, and about $700 million a year of that money was flowing back out to the contractors it calls fellows. No platform we have reviewed publishes numbers close to that.

Getting there at that speed meant onboarding 750,000 people with almost no humans in the loop. A timer decides how many of your hours get paid. An automated review decides whether your account survives. Until this summer there was no appeal. Two contractors sued over withheld pay, Business Insider found four of five suspended contractors on OpenAI projects were never paid for finished work, and in July 2026 a proposed class action landed in the Northern District of California.

So yes, Handshake AI is legit. The money is real and it arrives on Wednesdays. What this Handshake AI review is about is the gap between the rate on the listing and the rate you keep, who is allowed in, and the four things that get accounts closed with the pay still inside them.

Disclosure: Handshake AI's referral programme pays only its own fellows, so Skillora earns nothing from any link on this page. Some other platforms mentioned here run referral programmes; where they do, Skillora may earn a fee if you sign up through our link. It costs you nothing and has no effect on what we recommend.

Is Handshake AI legit?

What checks out:

• A real company with public numbers. Handshake AI is run by Stryder Corp., doing business as Handshake, the career network used by 1,600 universities. The AI unit launched in January 2025 and CEO Garrett Lord announced in October 2025 that the whole company was re-founding around it, laying off 100 people on the recruiting side to do so.
• Revenue that is audited by investors, not just claimed. Sacra puts gross annualized revenue at about $1.1 billion in April 2026, up 349% year on year, with roughly $450 million net after contractor payments. TIME named Lord to its 2026 TIME100 AI list on the strength of it.
• Eight frontier labs as customers. Handshake will not name them, but contractors interviewed by Business Insider were working on OpenAI projects.
• No fees, and paid onboarding. Nothing to apply, nothing to take the training. The roughly two hours of Canvas modules, quiz and practice task before your first project are paid, which most of this market does not do.
• Weekly pay on a published schedule. Monday to Sunday pay period, payout sent by Wednesday, through Stripe.
• A published visa list. Handshake states exactly which work authorisations it can and cannot support. Nobody else in this market does.

What to weigh against it:

• Two individual lawsuits and a proposed class action, covered below.
• Trustpilot at 1.9 out of 5 on the main domain, with 88% one-star reviews, and 1.2 on the app domain. Both mix career-network complaints with fellows' complaints, so read the topics, not the score.
• No appeal, until recently. Contractors suspended in early 2026 were told the decision was final. The help centre now lists a way to request one.
• You give a lot of identity data before you see any work. Phone, work authorisation, resume, then government ID, selfie, Social Security number and, on some projects, a background check.

Legitimate, huge, and run by systems that do not explain themselves. That is the honest position.

The number on the homepage is a year out of date

The Handshake AI homepage carries two badges: $100M payouts made and 750k+ fellows. The first is stale in a way that flatters the company. Handshake's own investor numbers put contractor payments at roughly $650 to $700 million a year at the April 2026 run rate, which makes the badge about seven weeks of payouts.

The second badge is stale in the other direction. 750,000 fellows is signups, not people who have been paid. Handshake's help centre is explicit that submitting an application does not guarantee placement and that not hearing back may simply mean no current project matches your background. There is no queue to browse and no timeline. Early-2025 reporting had 100,000 people paid out of a much smaller pool, and nothing published since gives a current ratio.

Both badges matter because they set your expectations wrong in opposite directions. The money is much bigger than the homepage says. Your odds of touching it are much lower than the fellows number implies.

How Handshake AI works: what the signup actually looks like

We went through the public flow ourselves and stopped at the account wall, since the rest requires creating a Handshake account.

The public board lists open projects with a ceiling on each card. Click Apply on any of them and you land on a login page, with Google or an email address, and a line saying you agree to the Terms of Service and Privacy Policy. You see no project description, rate band or hours until you have an account. Every listing link carries a job ID, so the application does record which role you clicked.

From Handshake's own onboarding guides, the rest of the path is:

1. Create or link a Handshake account. A student account is required; you can make one at this step.
2. Verify your phone with a one-time code.
3. Answer three work-authorisation questions: whether you can legally work in the US, whether you need CPT or OPT sponsorship, and whether you need H-1B sponsorship.
4. Describe your background and upload a resume. PDF only, up to 50 MB.
5. Wait to be matched. The guide says there is nothing else to do. You can also submit interest forms for specific roles from the Opportunities tab, which the post-application guide recommends.
6. On invitation, verify identity through Persona. Driver's licence, state ID, passport or permanent resident card, then a selfie sequence. One-time, before your first project.
7. Set up Stripe for payouts. Some projects then require a background check through Yardstik, which needs your Social Security number and usually returns within seven days. You can keep onboarding while it runs.
8. Complete project onboarding. Around two hours of Canvas learning modules, a knowledge quiz and a practice task with reviewer feedback. This is paid.
9. Pass any project assessment. Format and pass mark vary by lab. Retakes are, in Handshake's words, generally not available, though failing one project does not close your account.
10. Task with the timer running. Every task has a maximum handling time. Anything beyond it is unpaid.

Notice what is missing. There is no AI interview. Mercor, micro1 and Alignerr all screen with a live or recorded AI conversation that scores how you reason, as the platform screening guide sets out. Handshake screens on paper, then on your first tasks. That makes the door easier and the exit faster, which is the trade this whole review is about.

Who can join: the visa list nobody else prints

Handshake AI is US-only, and it defines US-only more precisely than any competitor. You need to be based in the US with a Social Security number or ITIN, and work authorisation that survives a contractor model rather than an employer one.

Supported: US citizens and permanent residents, F-1 students on CPT or OPT, H-4 spouses holding an EAD, L-2 spouses, DACA recipients, refugees, asylum seekers and grantees, and TPS holders.

Not supported: F-1 STEM OPT extension, J-1, H-1B, TN, and H-4 without an EAD. Handshake's stated reason is that most of these require an employer-employee relationship, and it operates under a contractor model.

The STEM OPT line is the one to reread if you are an international graduate student, because it excludes the largest group of them. Ordinary OPT qualifies. The two-year STEM extension, which requires a training plan with a named employer, does not. If you are on F-1 you also need your Designated School Official's approval before accepting a project invitation, and Handshake cannot provide a DSO letter until after you have been assigned to one. Plan for that sequence.

Two more geography rules from the help centre. Work has to be performed inside the US and its territories, and a US fellow away for 14 or more consecutive days needs written approval from a project lead before working remotely. Non-US residents can only join by invitation, though the public board now shows role variants marked India, Canada, EU and LATAM, so the invitation-only pool is real and growing.

What Handshake AI pays

Handshake publishes a ceiling on every listing, never a band. The words are always "Up to". Its guide for university partners gives the underlying range as $30 to $125 an hour, depending on education level, expertise and project, with typical commitments of 5 to 20 hours a week and projects running two weeks to three months or more.

We read every card on the public board, about 190 open roles across 13 pages at the time of writing. The ceilings sort into clear tiers, and the degree filters on each card tell you who each tier is for.

• Up to $40: AI Evaluation Specialist, the generalist track. Open to associate's and bachelor's holders. Third-party reports put actual generalist evaluation around $17 an hour.
• Up to $65: Software Engineer and Adversarial Prompt Expert.
• Up to $75 to $85: the bulk of the board. Finance, energy, supply chain, construction, HR, engineering and most named professions. Physics Expert, Mathematics Expert and Computer Science Expert all sit at $75, bachelor's to postdoc.
• Up to $100 to $125: design and tooling. CAD drafters, PCB and 3D Slicer specialists, game developers, audio engineers, video editors, digital illustrators. Also electrical, mechanical and semiconductor engineering experts at $120, master's and above.
• Up to $150: Biology Expert and Chemistry Expert, both PhD or postdoc only, plus Legal Professionals, STEM Professor and Quantitative Finance Researcher.
• Up to $175: Investment Bankers.
• Up to $200: Physicians and Psychiatrists.
• Up to $300: Medical Professional, Ophthalmologists, Anesthesiologists, Neurologists.

Two things in that list are worth more than the numbers.

First, the same role is priced by country. Physics Expert is up to $75 in the US and up to $47 in the India variant. Electrical Engineering Expert is $120 in the US and $96 in Canada. STEM PhD is $48 in the EU listing and $30 in the LATAM one. Handshake is expanding by invitation, and the invited are paid on a local scale.

Second, the ceiling tracks the credential, not the task. A physics PhD and a physics bachelor's may grade the same model output. The card says $75 for both, and Handshake's partner guide says the rate within that ceiling is set by education level. This is a credential market, the opposite of micro1, which prices on a demonstrated-skill interview and where, as the micro1 review covers, a strong engineer without a graduate degree can reach $100 an hour and up.

Some projects pay per task rather than per hour. Handshake's earnings guide confirms both models exist, and third-party reporting puts specialist per-task rates around $120 to $140 for coding and design work.

Why the posted rate and the paid rate diverge

A fellow quoted on Handshake's own homepage says rates are posted upfront and never move mid-project. A Trustpilot reviewer from May 2026 says the advertised $17 an hour came out at about $7. Both can be right, because the rate is not what moves. The hours are.

Every task carries a maximum handling time. Handshake's contractor agreement says you are paid for your time on that task only up to that cap. The help centre adds that caps are set from the average time similar tasks took, that you must still submit a task once you hit the cap, and that anything beyond it will not be paid. The timer only counts while it is running, and you start and pause it yourself.

So your effective rate is the posted rate multiplied by the share of your working time that fits inside the caps. If a $75 task is capped at 40 minutes and takes you an hour, you were paid $50 an hour. Careful, thorough experts, exactly the people the rate is meant to attract, lose the most.

It gets worse than unpaid minutes. Repeated cap hits are one of the named triggers for a time-log review, alongside high idle ratios, unusual timing patterns, frequent manual adjustments and overlapping timestamps. The agreement says a fellow who repeatedly exceeds the cap may be removed from the project or the programme. Being slow is not a rate problem here. It is an account problem.

The defence is mechanical. Read the maximum handling time in the project terms before you accept. Run your first five tasks against it with the timer honest. If you are consistently over, leave the project before the review finds you, because the platform will not adjust the cap for you.

What went wrong in 2026: suspensions, two lawsuits and a class action

This is the section the content-farm reviews skip. It is also the section that matters most, because the pattern is enforcement, not fraud.

The suspensions. On 4 March 2026 Business Insider reported that five contractors working on OpenAI projects through Handshake had their accounts suspended between late December and January with no warning. Four were not paid for work already completed. The reasons Handshake gave, in support messages the outlet reviewed, were discrepancies in background credentials, task completion times three to four times above benchmark, and tasking from a non-US location. One contractor was told the decision was final and that there was no appeal process. Handshake declined to comment. Dozens of similar accounts appeared on Reddit, which we cannot verify individually.

The two individual lawsuits. In August 2025 a contractor sued over pay withheld on an accusation of using AI tools, at a time when the project reportedly allowed ChatGPT. A court ruled the contractor should receive $6,475 and the case settled in October. In November 2025 a second contractor sued for $9,600 in unpaid wages; that case was dismissed in mid-January 2026.

The class action. In July 2026 a proposed class action was filed in the US District Court for the Northern District of California against Stryder Corp., doing business as Handshake, by the firm Bryson Harris Suciu & DeMay. Law360 reported it on 22 July. The complaint alleges that Handshake misclassified thousands of college students and others as independent contractors while setting pay unilaterally, mandating its own platform, capping weekly billable hours, leaving onboarding and meetings unpaid, and retaliating against workers who raised pay concerns. Handshake had not filed a response as of early August, and nothing in it has been tested in court. It is an allegation, and one you should know exists before you sign.

The contract changed after it. Handshake's contractor agreement was last updated on 20 August 2026. It now states that Handshake may temporarily deactivate an account without notice to investigate a suspected violation, that either side can end the agreement on seven days' notice, that work product is assigned to Handshake, and that disputes go to binding individual arbitration under a class action waiver. You can opt out of the arbitration clause within 30 days of accepting, by certified mail. That opt-out window is the single most useful sentence in the document, and it closes fast.

The review pages. Trustpilot's page for the main domain sits at 1.9 out of 5 across 25 reviews, 88% of them one-star. The app domain is at 1.2 across 116, but that page mixes complaints about the career network's premium subscription with fellows' complaints, so the topics matter more than the score. Among reviews from May to August 2026 the fellow-specific themes are consistent: a hidden active-time timer, retroactive reductions to logged hours, bans with no stated reason, and accounts fully set up with no project ever arriving. One five-star review from late July praises timely payment, which is also true.

Set that against the other side of the ledger. Eight labs. A billion dollars of gross revenue. Roughly $700 million a year reaching contractors. The complaints are not that Handshake does not pay. They are that an automated system decides who gets paid, and that until this summer nobody could argue with it.

The four ways to lose your Handshake AI account (and the pay with it)

Every documented suspension falls into one of four buckets, and each has a defence.

1. A credential discrepancy. Handshake verifies identity, qualifications and work location, and reserves periodic re-verification. Flagged accounts get asked for a transcript, an institution or employer letter, or a portfolio. Keep your resume, your Handshake profile and your degree records saying exactly the same thing, down to dates and institution names, and keep the documents to hand.

2. Time three to four times above benchmark. Covered above. Know the cap, work inside it, leave projects where you cannot.

3. Working from outside the US. No VPN, no tasking from abroad, no exceptions. If you travel for 14 days or more, get written approval first. A non-US IP is the one violation that is trivial to detect and impossible to argue.

4. Using an LLM. Handshake's original-work policy prohibits AI tools for developing prompt ideas, generating or editing content, solving tasks, and even grammar or writing tools unless the project authorises them in writing. Prompts lifted from Khan Academy, Chegg, textbooks or other annotation platforms also count as non-original. If Handshake suspects LLM use it freezes all payments pending review; if confirmed, the work is unpaid and you are removed. You are required to keep drafts, notes and working materials for 90 days after delivery, and Handshake can ask for them. Keep them, and work in a document that shows its own history.

And one action that is not a defence but is worth doing: read the arbitration clause the day you accept the contractor agreement and decide whether to opt out within the 30-day window. It costs a certified letter.

Handshake AI jobs: what is listed right now

Because the public board is the only part of Handshake AI you can see without an account, it is worth knowing how to read it.

The Domain filter lists 27 fields, from biology, chemistry and physics through economics, education, finance, law, medicine, nursing and philosophy to music, performing arts and photography. The Degree filter splits associate's, bachelor's, master's, PhD and postdoc, and most professional cards accept bachelor's and above while the $150-and-up science cards are PhD and postdoc only.

The board's centre of gravity at the time of writing was not what the marketing suggests. The largest cluster was energy and industrial: drilling, reservoir and production engineers, power market and renewable analysts, plant performance and emissions specialists, all at $80 to $85. The second was supply chain and retail operations, from demand planners to store managers, at $75 to $80. The PhD science roles the fellowship was built on are still there, but they are a minority of the cards.

Two cards stand out as unusual for this market. Research Study Participant, paid and in person, at up to $40, and Intelligence Expert at up to $200 with a note welcoming bachelor's holders. Both suggest labs asking for kinds of human data that a task queue does not capture.

The cards refresh often, and titles are sometimes swapped onto the same job ID, so treat any one snapshot as a sample. The way to stay current is to submit interest forms from inside the dashboard, where the roles carry their actual descriptions and hours.

Handshake AI vs Mercor, micro1 and the rest

All the platforms in this market are ranked side by side in the best AI training jobs roundup. Where Handshake sits:

• Mercor is the direct competitor for credentialed professionals. It advertises a $112 average and its physician network posts $110 to $250 an hour, but it matches under one percent of registrations after a 20-minute AI interview. Handshake has no interview and far more volume, with lower ceilings for the same licence outside the $300 medical specialties. The Mercor review has the matching numbers.
• micro1 is the opposite design. It screens on a live technical interview and a proctored assessment, then prices on what you demonstrated rather than what you hold. Its US-open engineering and PhD tracks pay above Handshake's ceilings for the same profile, listed below.
• Alignerr has the widest project catalogue and the best generalist rates, with the caveat that it runs two payment models and only one pays for time. Read the Alignerr review before accepting a project.
• Outlier publishes rates by domain and country, roughly $15 to $30 for general rating and up to about $100 for legal and medical review, with a volume problem covered in the Outlier AI review.
• DataAnnotation is the closest thing to Handshake's $40 generalist track, at roughly $20 to $40, gated by one unpaid assessment with a single attempt. See the DataAnnotation review.
• Terac pays per approved research session at $100 to $350, which for a sub-hour study beats most of Handshake's ceilings.

The structural difference is the gate. Every other platform makes you prove something before you are in and then mostly leaves you alone. Handshake lets you in on paper and then measures every minute. Which you prefer depends on whether you would rather fail an interview once or be judged by a timer for as long as you stay.

What pays more than Handshake AI right now

Handshake's ceilings are real, but for a US professional they are not the top of this market. Skillora's AI training jobs board had 1,364 listings open to US applicants when we checked, aggregated from Mercor, micro1, Alignerr, Outlier, DataAnnotation and Terac, refreshed daily. Handshake AI is not one of its sources, because Handshake's listings sit behind the account wall. These were the live rates above Handshake's for the same profiles.

micro1, US-open, with published bands rather than ceilings:

• BigLaw attorneys across family, immigration, litigation, M&A, corporate, healthcare, civil rights and criminal practice at $140 to $400 an hour. Handshake's Lawyers card says up to $110 and Legal Professionals up to $150.
• PhD and Academic Expert, Data Scientist, Senior Software Engineer, Corporate Attorney, Senior Journalist, Investment and Finance Expert and Management Consultant, all at $245 to $280.
• Physics Expert at PhD or postdoc level at $100 to $200, with a further eight physics specialisations at $80 to $170. Handshake's Physics Expert is up to $75.
• AI and ML engineers at $70 to $200, and $160 to $300 for internal platforms work. Handshake's Software Engineer is up to $65.
• Oncology and haematology clinicians at $160 to $200, veteran mental health experts and developmental psychology PhDs at $100 to $200.
• GitHub and open-source contributors at $50 to $150, with no interview credential required beyond the work itself.

There were 361 micro1 listings open to US applicants, and every one of them sits behind the Zara interview described in the micro1 AI interview guide. That is the price of the higher band.

Elsewhere on the board: Alignerr's Machine Learning Evaluation Specialist at $200 to $400 and its Lean 4 proof engineers at $170 to $200; Mercor's physician network at $110 to $250, offensive-security researchers at $200 to $250 and disease-area clinicians at $150 to $230.

Where Handshake genuinely wins on rate is the $300 medical specialty ceiling, above Mercor's $250, and at the very bottom, where its $40 generalist track needs no assessment at all. Everywhere in between, a US professional with the same credential is being offered more on a platform that publishes a band and asks for an interview.

Every listing on the board links back to the marketplace that posted it. You apply and get paid by them directly, and browsing is free. There is a micro1 jobs page with the current tracks and published rates.

Who should apply to Handshake AI

Apply if you are: a US-based master's, PhD or postdoc in a listed science, especially biology or chemistry at the $150 ceiling; a US physician or medical specialist, where the $200 to $300 ceilings lead the market; a licensed professional in energy, finance or engineering who wants paid hourly work without an interview; or someone who has failed an AI interview elsewhere and wants a paper-screened door.

Prepare properly first if you are: thorough by temperament. Check the maximum handling time on your first project before you accept it, and treat the first five tasks as a test of whether you can work inside it.

Look elsewhere first if you are: on STEM OPT, J-1, H-1B or TN; outside the US without an invitation; unwilling to hand over ID, selfie and Social Security number before seeing any work; dependent on income this month rather than a place in a matching pool; or a strong engineer without a graduate credential, for whom micro1's interview-priced bands are simply higher.

Verdict

Handshake AI is legitimate, moving more money to contractors than any platform we have reviewed, and, for US physicians, medical specialists and science PhDs, the highest ceiling available without sitting an interview. That is a real position and it is worth applying for.

The cost is that you are managed by a system. A timer decides which of your hours are paid, an automated review decides whether your account continues, a class action is testing whether that arrangement is lawful, and the contract you accept on the way in sends any dispute to individual arbitration unless you opt out within 30 days. None of that is hidden. All of it is in Handshake's own help centre and contractor agreement, and almost nobody reads them.

Read them, work inside the cap, keep your drafts, stay in the country, and qualify on at least one platform that prices your skill instead of your degree. Then Handshake AI is a good place to be paid on Wednesdays.