Friday, September 18, 2026

Unpaid Take-Home Assignments: Are They Exploitative? A Fair Look at Both Sides

The Pudding Team

You made it past the resume screen. You did the phone call. Now there's an email asking you to spend "about three hours" building a marketing plan, a data model, a lesson plan, a design mockup or a sample memo. Unpaid.

If your first reaction was somewhere between a sigh and a rant, you're in large company. Search Reddit, LinkedIn or GitHub for "unpaid take-home" and you'll find thousands of posts calling the practice exploitative, insulting or a scam. Some of that anger is earned. Some of it is unreasonable.

You won't get validation here, and you won't get a lecture about being realistic. What follows is what's actually true about take-homes: the bargaining dynamics, the law, the ethics, the hirer's reasoning and the economics. Then it gives you a practical way to respond.

Take-homes are also no longer a tech-industry quirk. Marketers get asked for campaign briefs. Analysts get asked for decks. Teachers get asked for demo lessons. Writers get asked for sample pieces. Account managers get asked for pitches. This debate is now about how professionals in general get evaluated for work.

1. Bargaining power: why hirers can, and do, ask

Start with the uncomfortable part. A hirer can ask for unpaid work because, at the point of asking, they hold most of the leverage. They have one opening and dozens or hundreds of applicants. Any single candidate who declines is, from the hirer's seat, a rounding error.

That's simply how a market behaves when applicants vastly outnumber openings. Employers add steps to the hiring process because the cost of a bad hire is high and the cost of losing one candidate at the top of the funnel is low. In a Robert Half survey, 93% of hiring managers said their hiring process took longer in 2025 than two years earlier. Longer processes with more stages are the norm, and the take-home is one of those stages.

Ethics has nothing to do with any of it. The mechanism is scarcity. And scarcity runs in both directions.

Run the thought experiment at the far end. If Jensen Huang, Nvidia's co-founder and CEO, decided to go job hunting tomorrow, no company on earth would email him a three-hour exercise. He'd be the one setting terms, and the hirer would be the one auditioning. Nothing about him makes unpaid assessment wrong in his case and right in yours. The only thing that changed is which side is scarce and which side is substitutable.

That's the whole mechanism. Leverage in hiring moves with how rare your skill is, how verifiable your track record is without a test, how many comparable candidates are in the pool, and what the market is doing that quarter. A senior specialist with a public body of work and three inbound offers is closer to the Huang end than they think. A career-switcher with no portfolio in a flooded field is closer to the other.

That reframes the question worth asking. "Is it acceptable to ask for unpaid work?" gets you an argument. "Where am I standing on this curve, and what does that let me ask for?" gets you a plan.

Candidates aren't powerless, though, and the data shows it. Monster's 2024 Work Watch Report found 36% of respondents had dropped out of a hiring process after applying because they were asked to jump through hoops such as unpaid assignments or multiple interview rounds. Monster didn't separate the two hoops, so the number isn't a clean take-home dropout rate on its own. Greenhouse's survey found about half of job seekers strongly dislike take-home assignments (hint: unpaid), and some drop out when they get one. Every good candidate who walks is a real cost to the hirer, even if no single one is visible.

The hirer has more leverage, then, but not unlimited leverage. Leverage is the context for negotiation.

2. What the law actually says (and what it doesn't)

Disclaimer: this is general information, not legal advice. Employment law varies by country, state and even city, and it changes. If you have a specific situation, talk to an employment attorney in your jurisdiction.

In most jurisdictions, the rule of thumb is this: if an exercise is designed to assess your skills and the company gets no business value from the output, it can be unpaid.

First, the word "work." Most arguments about take-homes break down because the two sides are using that word to mean different things. In its broadest and most accurate sense, work is effort applied to produce something. By that definition a take-home is plainly work, and there is no point pretending otherwise.

What does not follow is that payment is automatically owed. "This is work" and "this must be paid" are two separate claims, and the second does not arrive free with the first. Unpaid work is everywhere and mostly uncontroversial: studying for a certification, interviewing, building a portfolio piece, writing a conference talk proposal, doing a reference call for a former colleague. Effort alone has never been the trigger.

What matters is what happens to the output. Two questions decide almost everything. Who ends up owning what was made, and what does the hirer take from it? In a well-formed assessment the candidate retains ownership of the artifact and the hirer takes only signal, meaning information about whether to hire. When the hirer takes the artifact itself and puts it to use, that is a different transaction, and everything below changes with it.

In the United States, the reasoning runs through the Fair Labor Standards Act. Pre-employment testing is generally not compensable because the person taking the test is an applicant, not an employee. The pay obligation kicks in when the test stops being an evaluation and starts producing something the business uses. A written aptitude exam or a typing check almost never has to be paid. A trial shift where a candidate rings up real customers or cooks food that gets served almost always does.

The related trainee and intern doctrines point the same direction. Courts balance the benefit to the trainee against any advantage to the employer. When an employer is clearly gaining an economic advantage, the trainees will likely be deemed employees entitled to pay. Many other jurisdictions (the UK, Australia, much of the EU) ask a similar question, with different labels and thresholds. Is this genuinely an assessment, or is it work?

The nuance: "business value" vs. "commercial value"

What trips up both sides is that the law was never written with a careful line between "generalized benefit" and "commercial benefit." It reaches for that line with vague proxies, and the proxies have to be read in context.

In the US, the entire foundation is one sentence from a 1947 Supreme Court case about railroad brakemen trainees. In Walling v. Portland Terminal, the Court held that because the railroad received no "immediate advantage" from the trainees' work, they weren't employees under the FLSA. Legal scholars have noted the Court discussed "immediate advantage" in a single sentence, acknowledged the concept could be abused, and then declined to say more, which is a large part of why unpaid-labor questions remain confusing seventy years later. Modern courts have mostly moved on to a "primary beneficiary" test, and some circuits have been openly dismissive of the "immediate advantage" language altogether.

The UK is similar. The concept of a work trial is not defined anywhere in statute. The government's minimum-wage guidance lists factors such as whether the tasks have a value to the employer beyond testing the individual, and HMRC states the guidance is not binding or determinative in any case, with tribunals deciding case by case.

So what does "advantage" or "value beyond testing" actually mean? Read literally, de-risking a hiring decision is an advantage to the company, and that's the whole point of assessing candidates. But that reading would make every interview compensable, and no court applies it. What the legal standards are reaching for is commercial value: does the output enter the company's production-to-sales cycle? Does it ship, get published, get sent to a client, displace work a paid employee would otherwise do? That's the distinction the "business value" language is clumsily pointing at, and it's the one that decides cases in practice.

Since most hirers don't deploy take-home outputs commercially, most unpaid take-homes sit inside the legal bounds. That's the reality candidates should know before threatening a wage claim over a three-hour exercise.

It cuts the other way, too. If a company asks you to write a real blog post they then publish, design a landing page they then ship, or analyze their actual customer data to make a real decision, that's no longer an assessment. The hirer is taking the artifact rather than the signal, and in most places that's compensable regardless of what the email called it.

3. The ethics: small exercises, big multipliers

Legal and ethical aren't the same thing, and the design profession figured this out decades before the rest of us.

AIGA, the professional association for design, has held a long-standing position against speculative work, meaning uncompensated work done in hopes of getting paid later. AIGA's view is that clients risk compromised quality, since little time, energy and thought can go into speculative work, and designers risk being taken advantage of, because some clients see it as a way to get free work and it diminishes the true economic value of what designers contribute. AIGA's recommended alternative is to ask designers for examples of previous work plus a statement of how they'd approach the project, then put the selected designer under contract before real work begins.

The AIGA position is aimed at client work more than job interviews, but the logic transfers cleanly, and it exposes something the "it's only an hour" argument misses.

Say a company asks 40 candidates to each do a one-hour exercise. Each individual ask is small. But the company has now extracted 40 hours of skilled professional time, unpaid, for its own decision-making benefit. Thirty-nine of those people will get nothing. The ethical weight doesn't stay at "one hour." It scales with the number of candidates. A one-hour ask sent to 40 people is a one-week ask sent to one person, and nobody would call that trivial.

This is the strongest version of the candidate's argument, and hirers should sit with it: the smallness of an individual exercise does not make the aggregate practice small.

4. The hirer's case for a work sample

The take-home predates every current complaint about it. Software engineering has used coding exercises and take-home projects for decades, for a reason peculiar to that field: code is unusually easy to inspect. A stranger's submission can be read in ten minutes and it tells you something real. Design arrived at the same instinct through the portfolio review. Neither practice has anything to do with the job market of the last three years.

The case for a work sample is one sentence long. Every other stage of hiring evaluates a claim about the work, and a sample evaluates the work. A resume is a claim the candidate writes. An interview is a performance about those claims. A reference is a secondhand claim from someone with a reason to be generous. The sample is the only point in the process where a hirer looks at the thing itself rather than a representation of it, which is why the practice keeps surviving criticism it often deserves.

Where it earns its keep is the tie. By the time a search is down to three or four finalists, those people have comparable backgrounds and all of them interviewed well, which is how they became finalists. The resume has stopped separating them and so has the interview. Without a sample, the tiebreaker is gut feel, which is where bias operates and where "culture fit" does its work as cover. A comparable artifact, produced from the same brief and scored against the same criteria, is the one thing at that stage that yields a ranking a hiring committee can defend out loud.

That leads to a fairness argument candidates rarely hear. Greenhouse found that while half of respondents disliked take-home tests, candidates from historically underrepresented groups were more likely to favor them, and an analysis of over 300,000 submissions showed pass rates rose by roughly 7 to 10% when grading was anonymized.

That second figure cuts both ways, and hirers should be made to hold both halves at once. It shows that take-home grading is biased as ordinarily practiced. It also shows the bias largely disappears when the names come off, which no other stage of hiring can claim.

What changed recently is reach and weight. The format escaped technical hiring into marketing, operations, finance, teaching and account management, and inside technical hiring it moved from one stage among several to the stage that frequently decides. The cause sits upstream. Generative AI made applying nearly free, and the screening layer collapsed under the volume: LinkedIn now processes roughly 11,000 applications a minute, up 45% in a single year, and Greenhouse's 2026 benchmarks put applications per job at 244 and applications per recruiter at 746. When every application reads as fluent and keyword-matched, fluency stops distinguishing anyone, and 39% of candidates told Gartner they already use AI in the process. Identity went next: 6% admitted to interview fraud, and Gartner projects one in four candidate profiles will be fake by 2028. Recruiters responded by adding friction on purpose. Greenhouse's Ophir Samson told WIRED that teams which once wanted applying to be effortless were suddenly fielding 2,000 applicants in 24 hours and now ask for friction back.

That's why an old practice became load-bearing. It also explains why it landed on professions that never used it. A marketing team and an engineering team now face the same broken resume screen, and a work sample is the cheapest instrument either of them can reach for.

The limitation the industry keeps ignoring. A take-home shows the output of someone's thinking, which is a different thing from showing the thinking. Now that the output can be generated in minutes, the artifact alone proves less every month. Reasoning becomes visible only when a candidate has to account for the work: what they chose, what they rejected, what they would do differently with another week. Most take-homes never ask, which is the largest unforced error in the format as it is usually run. It also leaves friction as the actual product, paid for entirely by the person with the least power to refuse it.

5. The AI problem: why take-homes are changing shape

The same tooling that broke the resume screen is now working on the take-home itself. A deliverable that used to take three hours can be generated in three minutes. Hirers know this. Candidates know hirers know this. The result is a quiet arms race: longer, weirder, more "un-Googleable" assignments on one side, and increasingly polished, increasingly interchangeable submissions on the other.

This matters for the unpaid debate in two ways.

First, it weakens the classic unpaid take-home. If a deliverable can be produced by anyone with a subscription, it no longer discriminates between candidates, so asking 40 people to produce one unpaid is now both ethically heavy and informationally light. The worst of both worlds.

Second, it strengthens a different format: make the thing, then explain it. The signal has moved from the artifact to the reasoning behind it. "Show me your Jira setup and walk our reviewer through why you structured it that way" is very hard to fake, because the explanation exposes whether the candidate understands the choices or merely generated them.

6. The economics: where paying does and doesn't scale

If work samples are so useful, why not just pay for them? Plenty of candidates say exactly that: every assignment, every stage, paid.

At the top of the funnel, that math breaks. A screening exercise sent to 200 applicants at a defensible rate is a five-figure line item spent mostly on people the resume screen would have removed anyway, and it has to be re-spent on every open role. Early-funnel paid take-homes don't scale, and arguing otherwise won't persuade anyone who has run a hiring budget.

Later in the funnel, the math flips completely. By the time you're down to semifinalists and finalists, the numbers are small and the stakes are high. For roughly 2% of the role's annual salary, a hirer can pay a real honorarium to ten finalists. On a $90,000 role that's about $1,800 total, or $180 per candidate. Against a cost-per-hire that routinely runs into the thousands, and against the cost of a mis-hire, that's a rounding error in the other direction.

The reasonable position lands well short of "all take-homes must be paid": unpaid at the screening stage is defensible if the exercise is genuinely tiny, unpaid for finalists doing multi-hour work is hard to justify when the fix costs 2% of salary.

7. "Honorarium" vs. "compensation": why the word matters

Candidates usually say "pay me for my work." Hirers usually hear "you're claiming I owe you wages." That mismatch causes a lot of unnecessary friction.

The precise answer starts with what the hirer actually receives. In a well-formed assessment, they acquire nothing commercially deployable. The candidate keeps the artifact, keeps ownership of it, and can reuse it tomorrow. The hirer walks away holding information, and information about whether to hire someone has no resale value. So the hirer owes nothing for the artifact, because at no point did they buy it.

What they're paying for is the demonstration. The candidate spent time producing something and then presenting it for review, and that is the exact shape of an honorarium: a payment offered in recognition of someone's time and presentation where no purchase has taken place. It's the same instrument a conference uses for a guest speaker, who keeps their talk, their slides and their material, and is paid for showing up and delivering it.

That's also why the word does operational work. "Compensation" implies the company bought something and owes a rate for it, which drags in questions about scope, ownership, and what else the fee entitles them to. "Honorarium" carries none of that freight. The money changes hands and the ownership does not.

None of this turns on whether the candidate is a contractor. On a platform that pays them properly they are one, with the tax paperwork to match. Contractor describes their status. Honorarium describes what the payment is for. The two answer different questions, and confusing them is what makes these conversations awkward on both sides.

If you're a candidate asking for payment, "honorarium" is the word most likely to get you one.

8. How much is fair? A benchmark for honorariums

No industry standard exists yet. A defensible one, with an actual floor under it: an honorarium should clear the hourly rate of the job itself. If a role pays $90,000, it implies roughly $43 an hour. Paying a finalist less than that for work meant to demonstrate they can do the job is an insult with a spreadsheet attached.

That floor lands at about 0.2% of annual salary per finalist for a 1 to 3 hour exercise, the same 2%-across-ten-finalists budget as above.

Role salary (USD/yr)	Implied hourly	1 to 3 hour exercise	4 to 6 hour exercise	7+ hours
$50,000	~$24	$100	$200	Scope it down
$90,000	~$43	$180	$360	Scope it down
$150,000	~$72	$300	$600	Scope it down
$250,000	~$120	$500	$1,000	Scope it down

Three notes on reading the table. The amounts are flat rather than hourly, and they're set so that even at the top of each time band the effective rate stays above the role's own implied hourly. That's what makes them defensible to a candidate rather than merely non-zero. A 4 to 6 hour exercise doubles the per-candidate figure, so a ten-finalist round costs about 4% of salary. That's the price of asking for more, and it's a useful check on scope. And the "7+ hours" column is deliberately blank for two reasons. Payments that size start dragging in tax paperwork: the federal 1099-NEC threshold rose to $2,000 for 2026 payments, but most states kept theirs at $600, so W-9s and filings arrive well before the federal number does, which is a lot of overhead for a hiring exercise. The second reason matters more. Work of that weight is substantial enough to be useful, and useful work gets absorbed into production.

For candidates, this gives you a number to ask for and a reason it's the right number. For hirers, it's a line item that's trivial next to the recruiter fee, and one that no finalist will read as a token.

9. Candidate objections, taken seriously

"My resume should be enough. Or the interview."

Both signals are weaker than they feel, for the reasons above. There's also a second problem that candidates rarely see from their seat: ties. In a competitive search, several finalists routinely have comparable resumes and interview comparably well. The hirer needs a tiebreaker, and a work sample is a far better tiebreaker than gut feel, which is where bias lives.

Live interviews also don't scale. They're synchronous, so every hour of candidate time costs an hour of interviewer time, plus scheduling. One survey found about 42% of candidates withdrew from a process because scheduling took too long. An asynchronous exercise sidesteps that entirely, for both sides.

"Take-homes aren't reusable. I do all this work and it goes nowhere."

This one's correct, and the industry should admit it. Most take-homes are designed as throwaway artifacts: bespoke to one company, useless anywhere else, and owned by nobody in particular.

It doesn't have to be that way. A well-designed take-home is realistic and repurposable for the candidate, yet of no commercial value to the hirer. That's a narrow target, but it's hittable: an exercise adjacent to the company's domain, drawn from public or synthetic inputs, that produces something a candidate could legitimately put in a portfolio or reuse for the next application.

This is changing, and the change is being led by platforms that let candidates retain ownership of what they make. That's much harder to guarantee when a company collects deliverables directly by email, with no terms in place and no neutral medium to govern the exchange.

"I'm too busy to do take-homes."

Also valid. The people hirers most want are usually employed, often with families, and a "please return this in 48 hours" email is a filter for who has free time, not who's best.

The fix is straightforward: windows measured in weeks, not days, and an honorarium at a rate that signals respect. If a hirer can't offer either, that tells you something about how they'll treat your time once you're on payroll.

10. Hirer objections, taken seriously

Four objections come up repeatedly on the hiring side. Each has a straightforward answer.

"I don't have time to review take-home deliverables."

Work samples deliver signal fast. A reviewer with domain expertise often knows within a minute whether a deliverable is in the top tier or not. The remaining time goes to ranking the good ones. That's far less reviewer time per candidate than a 45-minute interview, and it can be done whenever the reviewer has free time.

"I don't want to insult my finalists."

Then don't. Offer an honorarium and a flexible timeline. Framed that way, finalists don't experience the exercise as an insult. They experience it as a process that judges them on their work rather than on how they performed on a Tuesday afternoon Zoom.

"I don't know if HR will allow it."

This is a real concern, and it's mostly about liability: does paying an applicant create an employment or contractor relationship? Does collecting their work create IP ambiguity? The clean answer is a third-party platform that sets the roles and terms in advance, limits the hirer's commercial use of the output, pays the honorarium automatically, and runs the exchange in a context that is explicitly not an employer-employee dynamic. HR can say yes to a defined process far more easily than to an ad-hoc one.

"I don't know what to ask them to do."

It's easier than it looks. A good exercise is something adjacent to your business that (a) lets candidates show their thinking and tool proficiency, (b) is not derived from any of your internal information, and (c) is not a useful input to your production workflows.

For example: "Set up a Jira project for a fictional product launch in our industry, and walk our reviewer through your structure and why you chose it." That gets you 30 minutes of the most honest signal you'll see in the whole process.

11. Red flags: when a take-home is really free labor

Most take-homes are clumsy, not malicious. But some are disguised production, and you should be able to tell the difference in about a minute. Any one of these is a yellow flag. Two or more, and you should assume the exercise has commercial value and treat it accordingly.

It uses the company's real data. Actual customer lists, real financials, live product analytics. An assessment can be built on public or synthetic data. A project can't.
It's a deliverable they visibly don't have yet. "Write our Q4 content calendar." "Design the onboarding flow for our new feature." "Audit our current pricing page and recommend changes." If the output would fill a gap on their roadmap, they're taking more than signal from it.
There are no ownership or usage terms. Nothing in writing says the company won't use what you submit. Silence defaults to whatever turns out to be convenient for them later.
Multiple rounds, each with a new deliverable. One exercise is a filter. Three is a workstream.
The scope is a full workday or more. Past six or seven hours, an exercise produces more output than an assessment needs. That is often just poor scoping, but it's fair to ask what the extra hours are meant to show.
The window is 48 hours or less on a multi-hour task. This one's about respect rather than exploitation, but it usually travels with the others.
They won't answer "how will this be used?" A hirer with an assessment answers immediately. A hirer with a project changes the subject.

None of this means you should refuse on the spot. It means you now have a specific, reasonable basis for asking for terms, a longer window, or an honorarium, and a specific reason to walk if the answer is no.

12. So what should you do when you get one?

The right response depends on the stage of the process, the scope of the ask, and what the output is for.

Ask three questions before you decide:

Will the output be used commercially? (If yes, the hirer is taking more than signal, and payment stops being a courtesy.)
How many candidates are doing it, and how long is it really? (Scale is what turns a small ask into a large one.)
Is there an honorarium, and how wide is the window?

If it's a small, early-stage screen with a generous window and nothing commercially useful in it, doing it is a reasonable bet, and often a better bet than another round of interviews.

If it's a multi-hour finalist exercise with no honorarium, you can push back without torching the relationship. Something like:

"I'm glad to do this, and I'd like to do it well. Given the scope, I'd ask for a modest honorarium and a two-week window. If it's easier on your side, I'm also happy to do it through a platform like Pudding that handles the terms and payment and lets me keep my work."

That's a proposal, and most hiring managers will hear it as one. A good hirer will recognize it as the reasonable ask it is. A hirer who reacts badly to it has just given you information you'd otherwise have paid for in your first ninety days.

Sources

Survey and benchmark figures throughout are cited via the outlets that reported them rather than the original questionnaires. Where a primary release exists (Gartner), it is linked directly. Validity meta-analyses (Schmidt & Hunter 1998, Sackett et al. 2022) were considered and deliberately left out: the estimates are heavily correction-dependent, were revised downward by as much as .21 in 2022, and rest on studies of work that predates remote hiring.

Back to blog