Platform review
Handshake AI
Rate ceilings published up to $150/hr, and a documented pattern of suspensions leaving finished work unpaid.
Handshake AI is the most transparent platform in this category about what it advertises and the least forgiving about what happens if it decides you broke a rule. Every role page publishes a ceiling, from Up to $25/hr for an India-based software engineer to Up to $150/hr for a STEM professor listing, and payment runs on ordinary contractor rails through Stripe on a weekly cycle. The risk sits at the other end of the relationship. Business Insider reported in March 2026 on five contractors suspended without warning on OpenAI projects, four of whom were not paid for completed work, and Handshake's own policy documents make that outcome contractual rather than accidental: on a confirmed LLM violation the work is rejected, the time is unpaid, and you can be removed from the program. Take it if the money is upside on work you would find interesting anyway; skip it if you are counting on the money.
Key facts about Handshake AI
- Pay
- Up to $25–$150/hr ceilings, published per role
- Application
- Interview / vetted (per platform)
- Availability
- Global
- Specialties
Read this first
The things that change how you should read everything else on this page.
- Every Handshake pay figure is a ceiling, not a rate. The role pages say Up to $X/hr, depending on the project, and the Contractor Agreement puts the amount, type, and timing of payment in the project invitation. Do not plan against the ceiling.
- Business Insider carries most of the weight in the negatives here. Its March 4, 2026 report documents five contractors and two lawsuits, which establishes a pattern with a court ruling behind it. It is not evidence that most Handshake fellows go unpaid, and nobody has published a suspension rate.
- The litigation cuts both ways. Business Insider reported a small claims suit for $9,600 filed in November and dismissed in January, and a separate suit filed in August that ended with an October ruling that the contractor was owed $6,475. That second contractor said Handshake withheld pay over alleged AI use.
- The Contractor Agreement binds you to individual arbitration with a class action waiver, which removes the collective remedy that the withheld-pay pattern would otherwise call for. The same agreement gives you 30 days from the effective date to opt out by signed letter to Handshake's San Francisco legal address.
- Handshake's own two help articles disagree on when payout setup happens, and under one of them you can complete billable work before a Stripe account exists. Set up identity verification and payout at the first opportunity rather than waiting for the post-task prompt.
- Handshake's site shows “$100M payouts made” and “750k+ fellows” as two separate statistics. It does not say the payouts went to those fellows, does not define what a fellow is, and puts no date or method on either figure, so no per-fellow average can be read out of them.
- We earn a commission if you sign up through our link, which is a /move-program/referral URL. We fetched it on September 12, 2026 and it resolves to the Handshake AI Fellowship page this review covers, with the same published role ceilings. Handshake does not document MOVE as a distinct program anywhere on its AI pages or help center, so treat the path as a URL for the fellowship rather than a separate offer.
What the work is
Handshake is the university careers platform, and Handshake AI is the arm that pays people to train and evaluate AI models for its lab partners. It runs as the Handshake AI Fellowship: you apply, get matched to a project, take training, complete tasks, and get paid hourly, per task, or per milestone. The work is reviewing and improving model output using your own subject expertise, across fields that run from semiconductor design to philosophy. Business Insider reported in March 2026 that some of the work is on OpenAI projects.
Who they want
Handshake accepts associate's through PhD holders, postdocs, and professionals with domain expertise, and says no AI experience is required. Its help center states no minimum education level at all, so the degree language on the marketing page describes who the company courts rather than a bar it enforces, though its own top-rate headline pitches remote part-time roles specifically at PhDs and Masters. The eligibility rules on the marketing page and on the role board do not match: the fellowship page says you must be US-based with valid work authorization, that F-1 students eligible for CPT or OPT may qualify, and that STEM OPT is not supported, but at least 17 of the 221 listings on our board are explicitly non-US, covering India, Canada, the EU, and Latin America. Those listings pay less than their US equivalents and Handshake's tax article handles W-8 filers, so read the individual listing for its geography rather than the marketing page.
Step 1
Create the account
Email or Google, a phone code, three work-authorization questions, optional school.
Step 2
Profile and agreement
An expertise profile, a resume PDF under 50 MB, then the Contractor Agreement.
Step 3
Project match and interest email
Handshake sets no timeline: projects launch on rolling partner demand.
Step 4
Selection, onboarding, training
Assessments come with the project, and no pass standard is published for them.
Step 5
Payout setup, then a first task
Handshake's two help articles disagree on the order; set up Stripe as early as allowed.
How you get in
You create an account with email or Google, optionally link a school, verify your phone with a one-time code, answer three work-authorization questions, fill in an expertise profile, upload a resume as a PDF under 50 MB, and accept the Terms of Service, Privacy Policy, and Contractor Agreement. From there Handshake's own sequence is a project match and interest email, a decision and selection step, project onboarding, training, and then your first task. Projects launch on rolling partner demand and Handshake states there is no fixed timeline for receiving a project invitation, so the gap between being accepted and doing paid work is open-ended. The company's two help articles disagree on when you set up payment: the application guide says identity verification and payout setup are required before you can start project work, while the application-to-first-project guide says you are prompted to set up Stripe after completing your first task. Do the payout setup as early as the system allows, because under the second sequence you can finish billable work before any payment method exists. Assessments are part of project participation, and Handshake publishes no pass standard, no retry policy, and no appeal route for them.
How it pays
Priced as a maximum, not a range. Handshake AI prices 200 of its 200 open roles as a ceiling — "up to" a figure, with no bottom stated. A distribution needs both ends, so these are listed rather than charted.
Every role page publishes a ceiling, written as Up to $X/hr and qualified as depending on the project, not a range and not a rate. Ceilings verified on September 12, 2026 run from Up to $25/hr (Software Engineer, India) to Up to $150/hr (STEM Professor), above the site's own Up to $100/hr headline, and the marketing page's examples sit between Up to $40/hr for an AI Evaluation Specialist and Up to $125/hr for a Game Developer/Designer. The binding number arrives in writing with the project invitation, per the Contractor Agreement, which sets payment as hourly, per task, or per milestone and classes you as an independent contractor rather than an employee. Payouts run through Stripe, sent by Wednesday afternoon covering the previous Monday 12:00 am to Sunday 11:59 pm PST, with up to a further 48 hours of processing. Tax is W-9 or W-8, with a 1099-NEC where payments reach the $600 reporting threshold Handshake's tax article names, and the company can pause tasking or withhold payment if no valid tax form is on file.
What workers say
The public record on Handshake AI
Independent material on Handshake AI specifically is thin in volume and severe in content. The two Trustpilot pages measure different things: the handshake.ai page carries 2 reviews and a score that does not reconcile with its own star breakdown, while joinhandshake.com's 1.9 from 26 reviews covers the student careers product and the AI fellowship together, so neither is a clean read on the fellowship. The weight sits with Business Insider's March 4, 2026 report, which documented five contractors, none named in the piece, two lawsuits, and dozens of others who the article says have shared similar stories on Reddit. Two of the 26 joinhandshake.com reviews are five-star, and both praise payment reliability and the efficiency of the process.
Scores, and what each one measures
- Trustpilot (handshake.ai) ↗2.9 out of 5
What it measures. Two one-star reviews of the AI fellowship, and nothing else. The displayed star breakdown is 100% one-star, which cannot produce a 2.9 score, so the score is not usable and the reviews are the only content worth reading. The August 2026 review describes a permanent ban after accidental VPN use and notes that a partner in identical circumstances was reinstated on appeal while the reviewer was not; it also criticizes support responsiveness. The July 2026 review is a one-star with a hostile title and no substantive detail. Profile unclaimed, with no history of soliciting reviews.
- Trustpilot (joinhandshake.com) ↗1.9 out of 5
What it measures. The Handshake company as a whole rather than the AI fellowship. Reviews of the student careers platform sit alongside AI training complaints under one score, so the number describes a bundle. Breakdown is 84% one-star, 8% two-star, 8% five-star. The recurring AI-training themes are frozen accounts with unpaid balances, permanent bans for alleged LLM use without the flagged submission being shown, unpaid training and assessments, and advertised rates not matching what arrived. Profile claimed by the company.
Our read of it
Pros
- Every role page publishes a rate ceiling before you apply. Ceilings verified on September 12, 2026 ran from Up to $25/hr to Up to $150/hr, which is more pay information than most platforms in this category disclose at any stage.
- Payment rails are conventional and dated. Stripe, a weekly payout window with a stated cutoff, and published W-9/W-8 and 1099-NEC guidance down to the $600 threshold mean you can predict the banking and tax side before you start.
- Handshake puts its unpaid activities in writing before you sign. Its earnings article names account setup, optional meetings and office hours, community platform time, and time lost to technical issues as unpaid, so the unwelcome part is findable in advance rather than discovered on a payslip.
- The credential gate is wide. Handshake says no prior AI experience is required and lists associate's through PhD holders, postdocs, and domain professionals, and its help center sets no minimum education, so the application does not screen on AI-specific work history the way vetted marketplaces do.
- The arbitration clause has an opt-out. The Contractor Agreement lets you send a signed letter to Handshake's San Francisco legal address within 30 days of the effective date and keep your right to sue, which is the highest-value thing you can do in your first week on the platform.
- Contractors who get onto a project are paid real money through a normal contractor pipeline. Business Insider's March 2026 report describes people who had been working and earning before their accounts were suspended, and the dispute it documents is over withheld pay rather than over whether the platform pays.
Cons
- Suspension with completed work left unpaid is documented. Business Insider reported on March 4, 2026 that five contractors on OpenAI projects had accounts suspended without warning between late December and January, and that four of the five were not paid for work they had finished. Handshake declined to comment for the article.
- Handshake told those workers there was no route back. Business Insider quoted the company saying “This decision is final. There is no appeal process,” with the associated work ineligible for payment. The reasons cited were credential discrepancies, task times three to four times above benchmark, and work performed from non-US locations.
- The non-payment mechanism is in the contract you sign. Handshake's LLM policy bans AI tools for generating work product, including LLM-based grammar and formatting tools, requires you to keep drafts and notes for 90 days after delivery, suspends payment for the work while a suspicion is reviewed, and on a finding rejects the work, pays nothing including for the time spent, and allows removal from the project or from the program entirely. It gives you a chance to submit evidence and describes no appeal.
- The ceiling is not an offer. Every published figure is a maximum that depends on the project, and the number that binds you arrives only with the invitation. A testimonial on Handshake's fellowship page credited to Guga Gogia, PhD Physics, says rates “never move, never change,” while reviewers on the joinhandshake.com Trustpilot page describe hours retroactively cut in half by an Active Time timer they were not told about until after payday, and an advertised $17/hr project that paid “about seven bucks an hour before taxes.”
- The unpaid surface is wide and the paid surface has a lid on it. Setup, optional meetings, community platform time, and technical downtime are unpaid by policy, onboarding is paid only where the Project Terms say so, and where a task carries a maximum handling time your payment stops at that limit even if the task runs longer. Assessments are part of project participation and Handshake publishes no standard for what they test.
- Project flow is unguaranteed and the exit terms run one way. Handshake states there is no fixed timeline for a project invitation. Either side can end the agreement on seven days' written notice, and Handshake can end it immediately when it determines a material breach occurred, on its own finding. Its offboarding article covers voluntary exit only.
Who it fits, and who should skip it
Best for
- Specialists with a credential in a field on the board who want a published ceiling before applying
- People treating the income as upside rather than rent money
- Anyone whose workflow involves no AI writing, grammar, or formatting tools
- US-based applicants, and non-US applicants willing to work at a lower ceiling
Who should skip it
Skip it if you need the money to arrive on a schedule you can plan around. There is no committed timeline to a first project, no published assessment standard, and a documented pattern of accounts suspended with finished work going unpaid and no appeal. Skip it too if AI writing or grammar tools are part of how you work, because Handshake's policy treats that as grounds to reject the work, withhold payment for the time spent on it, and remove you from the program.
Provenance
Where all of this came from
Every source behind this review
What each one says, and how much weight it deserves.
Earn up to $100/hr headline; example roles from Up to $40/hr to Up to $125/hr; US-based with valid work authorization, F-1 CPT/OPT may qualify, STEM OPT not supported; associate's through PhD plus postdocs and domain professionals, no AI experience required; $100M payouts made and 750k+ fellows shown as separate stats; Guga Gogia PhD Physics testimonial on rates never moving.
How to read it. Marketing page. The rate-stability claim is a named fellow's testimonial, not a commitment Handshake makes in its own voice, and the eligibility text is narrower than the role board.
Up to $150/hr, depending on the project. Above the site's own Up to $100/hr headline. US-based faculty with valid employment authorization.
How to read it. A ceiling, not an offer. Re-fetched September 12, 2026.
Up to $30/hr, depending on the project, for candidates residing in Latin America. Confirms non-US listings exist and pay less.
How to read it. Contradicts the US-only eligibility text on the fellowship marketing page.
Up to $25/hr for India-based applicants, the lowest ceiling verified on the board.
How to read it. Same role family as US software engineering listings at higher ceilings.
Task prep, submission, and required revisions are paid. Account and payment setup, optional meetings and office hours, community platform time, and time lost to technical issues are not. Onboarding is paid only where Project Terms say so, and payment stops at the maximum handling time where a cap applies.
How to read it. None. Handshake states this plainly in its own help center.
Sequence is application, project match and interest email, decision and selection, onboarding, training, then payout setup. No fixed timeline for a project invitation because projects launch on rolling partner demand. Stripe is set up after the first task.
How to read it. Conflicts with the application guide, which says payout setup is required before project work.
Account creation, optional school link, phone verification, work authorization questions, expertise profile, resume as a PDF under 50 MB, and acceptance of the Terms of Service, Privacy Policy, and Contractor Agreement. Identity verification and payout setup required before project work. No minimum education stated.
How to read it. The no-minimum-education point is the operative one; the marketing page's degree list is not a stated bar.
LLM-based grammar, style, and formatting tools are prohibited alongside generation tools. Drafts and notes must be kept 90 days after delivery. Payment is suspended on suspicion, and on a confirmed violation the work is rejected with no pay including for time spent, and the fellow may be removed from the project or the program.
How to read it. An opportunity to provide evidence is offered; no appeal process is described.
Independent contractor, not employment. Payment amount, type, and timing are communicated in writing with each project invitation, as task-based, time-based, or milestone-based. Either side may terminate on seven days' written notice, and Handshake may terminate immediately on a material breach it determines. Binding individual arbitration with a class action waiver, with a 30-day opt-out by signed letter.
How to read it. The 30-day opt-out window runs from the effective date and is easy to miss.
Payouts sent by Wednesday afternoon covering Monday 12:00 am to Sunday 11:59 pm PST of the previous week, arriving within about 48 hours depending on the bank. Stripe is the payout platform.
How to read it. None.
W-9 for US persons, W-8 for non-US persons. Reporting threshold for US independent contractors is $600 or more, though not every W-9 filer receives a 1099. Missing tax forms can lead to paused tasking or withheld payments.
How to read it. The W-8 handling corroborates that non-US fellows are paid through the same system.
Five contractors on OpenAI projects suspended without warning between late December and January, four unpaid for completed work. Cited reasons were credential discrepancies, task times three to four times over benchmark, and non-US work. Handshake told workers the decision was final with no appeal. A $9,600 small claims suit filed in November was dismissed in January; a suit filed in August ended with an October ruling that the contractor was owed $6,475. Handshake declined to comment.
How to read it. We read the AOL syndication of the Business Insider piece rather than the original. Five contractors is a documented pattern, not a measured rate.
2.9 out of 5 from 2 reviews on September 12, 2026, with a breakdown showing 100% one-star. Both reviews are one-star; the August 2026 one describes a permanent ban after accidental VPN use with a partner reinstated on appeal in identical circumstances.
How to read it. The score and the star breakdown do not reconcile, so the score is not usable. Two reviews cannot carry a verdict. Profile unclaimed.
1.9 out of 5 from 26 reviews on September 12, 2026, 84% one-star and 8% five-star. Themes include frozen accounts with unpaid balances, bans for alleged LLM use, unpaid training and assessments, an undisclosed Active Time timer cutting logged hours, and advertised rates not matching actual pay.
How to read it. Covers the student careers platform and the AI fellowship under one score, so it measures the company rather than the work.
Mirrors a Handshake AI legal and administrative trainer role at an estimated $75/hr, requiring at least four years in legal or administrative support, asynchronous and remote.
How to read it. Not independent corroboration. The figure is excerpted from Handshake's own posting and the page links straight to Handshake's ATS listing; $75/hr also appears as a ceiling on Handshake's own role pages.
Evidence scorecard
What we can back up about Handshake AI
15
0
Not one of them is work we've done ourselves. Read every line below as what somebody else states, discounted by its tier.
- Tier 1 · Verified first-hand0 accounts
We did it ourselves and can describe it, with a date. No dated first-hand account of Handshake AI's own process is on record here.
- Tier 2 · Platform-published11 statements
Stated on the platform's own site, board, or in writing to us. Read off Handshake AI's own pages, and cited to them.
- Tier 3 · Worker reports0 reports
Said by people who say they've done the work. We hold none for Handshake AI.
- Tier 4 · Unverified claim4 claims
Printed by a marketing or aggregator page with no sourcing behind it. We repeat these labeled so you can recognize them elsewhere, never as fact.
Listed pay
No rate publishedTier 2 · Platform-publishedNone of the 200 open roles we mirror from Handshake AI publishes an hourly rate. That's Handshake AI's own board, not a gap in our sync.
Effective pay
Not knownNobody has published itWhat a person clears per hour of their life after unpaid application time, unpaid training, idle time between tasks and self-employment tax. Nobody publishes it for Handshake AI, and we can't tell you yet either.
Getting in
Interview / vetted (per platform)Tier 2 · Platform-publishedHandshake AI's own material describes the step, and the review names the page it came from. What share of applicants clear it, Handshake AI does not publish.
Specialty labels
Our classificationOur inferenceHandshake AI publishes no specialty for at least some of these roles, so the labels you see on this site are our inference from role titles rather than Handshake AI's own taxonomy.