Somebody buys your answers before you ever see the survey. A research buyer pays a set price for each finished interview. That money walks down a chain: a panel, a router, then the app you're in. Each layer is paid by the one above it, none of them publishes what it keeps, and what reaches you is your reward. The same targeting that sets the price decides whether you get screened out. The buyer is paying for a certain kind of person, not for your time.
That's the answer. The rest of this page names those layers, explains why a long survey is priced the way it is, and shows why you can answer honestly and still be sent back with nothing.
Who actually pays for your survey
Five parties stand in a line, and money walks down it one step at a time.
At the top is the buyer. A brand, an agency or a university wants answers from a defined group, and it doesn't go out and find those people itself. It buys them.
Below that sits a panel or a sample exchange, such as Cint. In Cint's own pricing contract, the thing a buyer pays for is a Complete, defined as "a single instance of a completed Survey by a Participant recorded by the Cint Exchange but which has not been reversed under the Reconciliation Policy". Read the last eight words again, because a finished survey only counts for as long as it has not been taken back.
The people come from what the same contract calls a Supply Partner: "an entity that recruits and obtains consent from and/or enters into contracts with Participants".
Next comes a router. Firms like CPX Research, BitLabs and TapResearch place surveys inside apps, and they decide which survey you're shown.
Then the app, and then you.
Each step is paid by the step above it.
| Layer | What it does | What it takes |
|---|---|---|
| The buyer | Needs answers from a defined group of people | Pays a set price for each finished survey |
| The panel or exchange | Recruits people, and prices each survey on a rate card | Its share is not published |
| The router | Places surveys inside apps, and picks which one you see | Its share is not published |
| The app | Shows you the survey, and turns the payment into its own currency | The gap between the two, never published |
| You | Answer the questions | What's left, once the hold has passed |
Why a survey costs what it costs
The price isn't a guess: it comes off a rate card.
Cint's pricing terms call the Rate Card "the table maintained on the Cint Exchange that sets out the CPI payable for Completes". CPI is the cost per interview. The same contract says it's set by "factors such as Participant territory, length of interview and incidence rate".
Three are named, then: where you live, how long the survey runs, and how rare you are.
Length is the easy one: the contract names it as a factor, and Cint's developer docs add that "longer surveys can affect respondent willingness".
Territory is the quiet one. The same twenty minutes of your time is priced differently depending on which country the buyer needs the answers from, because the contract names the participant's territory as one of the factors.
The incidence rate is the part nobody explains. Cint's developer docs define it as "your estimated proportion of the target population that will qualify for the survey", and add that "a lower IR typically means higher costs or longer field times".
A survey hunting for dog owners who changed insurer last year has a low incidence. Few people qualify, so the buyer pays more for each one who does. A survey open to any adult has a high incidence and costs less.
Your reward follows that price. It doesn't follow your effort.
How low does the top of the chain start? SurveyMonkey sells answers from its own panel "starting at $1 USD per response", on its published product page, read on 22 September 2026. That's the buyer's price, at the top of the chain.
Why you get screened out of surveys in the first few minutes
The screener is the incidence rate doing its job.
The buyer has bought a specific group, and the first questions check if you belong to it. If you don't, the survey ends. You didn't fail anything: you're simply not who was ordered.
TapResearch, one of the routers, says it plainly in its own developer FAQ: "Many surveys are looking for very specific audiences and may disqualify users based on their answers to particular questions."
One detail explains the frustration better than anything else. The rate card prices Completes, and a screener that ends in nothing is not a Complete, so it never reaches that bill. That is why the minutes you spent there usually pay nothing.
Profiling happens before that as well. SurveyMonkey's published answers to the industry's standard questionnaire describe core profiling of "age, gender, US zip code, and household income range", with "over 75 profiled attributes" available in total. CPX Research marks some surveys as needing more profiling before the survey can start at all.
Why you can qualify for a survey and still be turned away
Matching the target isn't enough, because there also has to be room in your part of it.
Cint's glossary defines a quota as the number of respondents "available for each selected profile condition". A buyer doesn't want a plain 500 people, it wants 500 people spread across set groups. SurveyMonkey says that by default it balances "on 4 buckets of age and gender".
Once your bucket is full, you're out. You matched, you answered, and the slot was already gone.
That's why screening can feel random.
The three ways a survey ends without paying you in full
| What happened | What was decided | When it happens | Did you do anything wrong |
|---|---|---|---|
| Screened out | You're not in the group being bought | In the first minutes | No |
| Quota full | You match, but your slot is taken | Any time | No |
| Rejected in checking | Your answers failed a quality test | After you finish | Sometimes |
Only the third one is about you, and even then not always. TapResearch's publisher docs list why researchers reject a finished survey: "VPN/Proxy usage", "Gibberish or nonsensical answers on open-ended questions", "Speeding - Completing surveys too quickly to have read the questions", "Straightlining - Selecting answers in a straight line on grid questions", and "Inconsistent responses - Contradictory answers".
SurveyMonkey describes the same step from the panel's side. A flagged answer is caught by "an AI-powered model that detects fraudulent behaviors", and then "we automatically delete the response and the panelist is not paid for their submission".
That decision can also land after you finished. The same Cint contract defines its Reconciliation Policy as the policy "for reversals of Completes that do not meet certain quality standards", and Cint's developer glossary calls the act itself the "ability to change the respondent session status from complete to term or term to complete". BitLabs names "reconciled survey" as one of the event types its system sends to apps. For the money side of that, and what pending means on your balance, see why a balance is still pending.
What the app is paid for a survey, and what you're paid
These are two different numbers, and the routers' own docs show both.
BitLabs sends an app a callback for every finished survey. One field is the "User Reward in App Currency", and the docs add that "You can configure it in the Currency Settings". The next field is the "Complete payment in USD, paid to the publisher". One message, two amounts, and the app configures the one you see.
CPX Research documents the same split: one field is the "Payout to your user in your local currency (based on your currency settings in publisher profile)", and another is the "Survey payout to Publisher in USD".
Its documentation goes one step further, because the sample survey it prints has both numbers filled in. That example carries a publisher payout of 0.81 US dollars and a user payout of 0.40, in the same record, read on 22 September 2026. It is sample data in a manual, not a real payment, and the user number is whatever the app set in its own settings. The ratio proves nothing about any real app. What it shows is that two different amounts for the same survey is the normal shape of the data, plain enough to print in an example.
BitLabs also hands the app the raw price. Its API reference defines cpi as the "CPI of this survey in USD without any formatting applied", and value as that same "CPI formatted according to your app settings".
Read that again: the app receives the price in dollars and then decides what you see. What the app keeps is decided in those settings, and no rewards app publishes what it keeps on a real survey.
Two more fields are worth knowing about. BitLabs gives every survey a rating, defined as a "Difficulty ranking of this survey. 1-5 (1 = hard, 5 = easy)". It also sends a cr field, the "Conversion rate of the survey", as a value between 0 and 1. The app receives both numbers for every survey. Whether you ever see them is the app's choice.
What each layer of the survey chain says about its cut
Every figure below is the company's own published wording, read on 22 September 2026.
| Who | What it publishes about its own cut |
|---|---|
| Amazon Mechanical Turk | "20% fee on the reward and bonus amount (if any) you pay Workers", plus "an additional 20% fee" on tasks with ten or more assignments |
| Prolific | "For pay-as-you-go pricing, our platform fee is usually 42.8% for corporate customers, and a discounted 33.3% for academic or non-profit customers" |
| SurveyMonkey Audience | The buyer's price only: "starting at $1 USD per response" |
| Survey routers inside apps | Nothing |
| Rewards apps | Nothing |
The two platforms that print a percentage aren't rewards apps. Mechanical Turk is a task marketplace. Prolific is a research platform with a separate rate for academic and non-profit buyers, and it also sets a floor on what buyers must pay: "the minimum pay allowed is £6.00 / $8.00 per hour".
That floor is one company's policy on its own platform, not a market rule, and it isn't a figure anybody should read as what an app pays.
So what do surveys really pay?
We're not going to print a figure, and any page that does is describing one person on one day.
Three reasons, all of them checkable.
No router whose documents we read, and no rewards app, publishes what it keeps. The two platforms above that do publish a cut aren't in this market.
The number on a survey isn't your rate. Your real rate is that number divided by all the time you spent, including every screener that ended in nothing. Nobody can know that ratio but you.
Published claims don't agree with each other. Reading the pages that rank for this topic on 22 September 2026, the figures ran from about a dollar a day to a couple of hundred a month, and not one showed how it was measured.
What can be said with evidence is narrower and more useful. A screenout usually pays nothing. TapResearch, one of the routers, says it "was the first to create and patent a payout schema we named Guaranteed Payout", which "pays every user in proportion to their time spent answering questions". Read that as what it admits. Paying people for time they had already spent was new enough to patent, and one company claims it got there first. That tells you what the default was.
The app waits too: TapResearch's publisher docs state that "Payments are processed approximately 90 days after the month the revenue was earned". An app on TapResearch is paid months after you finish.
What actually changes your odds with surveys
None of this is a trick, and none of it earns you more per hour. It's the mechanism, used properly.
- Fill in your profile, fully and honestly. Matching runs on profile answers, and at one panel that's over 75 attributes. A thin profile means fewer matches.
- Answer the screener straight. Guessing what they want produces contradictions, and "Inconsistent responses" is a named reason for rejection.
- Don't rush. "Speeding" and "Straightlining" are both on the published rejection list.
- Don't use a VPN. It's the first item on TapResearch's list of why researchers reject answers.
- Expect more screenouts on the surveys that pay more. A narrow target means a low incidence rate, and Cint's docs say a lower one "typically means higher costs". So the surveys that pay more are often the ones fewest people qualify for.
That last one is the trade the ranking pages leave out: the best paying surveys are often the ones you're least likely to be in.
Which apps and networks run surveys
Surveys reach you in one of two shapes. In the first, an app doesn't run surveys itself: it plugs in one router or more.
The routers you'll meet inside apps include CPX Research, BitLabs and TapResearch.
In the second, you join a panel directly, such as Survey Junkie or Opinion Outpost, and the panel recruits you itself and sells your answers upward. Even a panel is not always one pool of people: SurveyMonkey says its own panel is two panels of its own "augmented by third party global partners".
One consequence is worth knowing. Each app sets its own rate on the dollars it receives, so the same kind of survey can pay differently from one app to the next.
How the survey lane works in PaysFair
We're the first kind. Surveys reach you through survey partners rather than a panel of our own, so the screening, the quotas and the checking above all belong to the buyer, not to us. We can't shorten a screener or free up a quota, and neither can any other app.
Three things we decided on our side, and they're in the code rather than in a promise.
Surveys are their own switch, off until you turn them on, and every way of earning in the app works the same way.
The survey walls stay shut on a VPN connection and say so plainly, because survey partners refuse that traffic. If our own check is down, you're let through rather than locked out.
Every earning shows its own release date, and the hold is set per partner. What pending means, and when money can still be taken back, is on our page about a balance that's still pending.
We don't publish what a survey pays here. The honest answer is the one above: it depends on the buyer's price, on the layers in between, and on how often you get screened out.
Sources
- Cint Exchange Rate Card Pricing Terms, version 2025/01, Cint. Published 1 December 2025, read 22 September 2026.
- How feasibility works, Cint Exchange Demand API documentation, version 18 December 2025. Read 22 September 2026.
- Glossary, Cint Exchange Demand API documentation, version 18 December 2025. Read 22 September 2026.
- Callbacks, BitLabs publisher documentation. Updated 11 August 2026, read 22 September 2026.
- Get Surveys V2, BitLabs publisher API reference. Updated 20 May 2026, read 22 September 2026.
- CPX Research publisher documentation. Read 22 September 2026.
- Developer FAQ for mobile app publishers, TapResearch. Read 22 September 2026.
- FAQ, TapResearch publisher documentation 3.x. Read 22 September 2026.
- Answers to the ESOMAR 37 questions, SurveyMonkey. Sample figures dated 2024, read 22 September 2026.
- SurveyMonkey Audience panel pricing, SurveyMonkey. Read 22 September 2026.
- Pricing, Amazon Mechanical Turk, Amazon Web Services. Read 22 September 2026.
- Pricing, Prolific. Read 22 September 2026.
- PaysFair's own backend source, read 22 September 2026.
We quote other companies in plain text and never link them. Rate cards, platform fees and router docs change without notice. We check every figure on this page again by 22 March 2027.
Questions people ask about paid surveys
Who actually pays for paid surveys?
A research buyer, such as a brand, an agency or a university, pays for each finished interview. That money passes through a panel or sample exchange, then a survey router, then the app you are using. Each layer is paid by the one above it, and what reaches you is your reward.
Why do I get screened out of surveys after a few questions?
Because the buyer paid for a specific group of people, and the first questions check if you belong to it. The industry calls that share the incidence rate. You did not fail anything. The rate card prices finished surveys only, so a screener that ends in nothing never reaches the buyer's bill.
Why do I get told a quota is full when I qualified?
A buyer does not want a plain headcount. It wants a set number of people in each group. One panel says it balances on four buckets of age and gender by default. Once your bucket is full you are turned away, even though you matched.
Why does a long survey pay so little?
The price is set by a rate card, not by effort. Cint's pricing terms say the cost per interview is set by factors such as the participant's territory, the length of the interview and the incidence rate. Rare people cost more. Common people cost less. Your reward follows that price, not your effort.
How much do surveys really pay per hour?
No honest number exists for this market. No survey router or rewards app we have read publishes what it keeps, and your real rate depends on how often you get screened out, which nobody can know but you. Any page quoting an hourly figure is describing one person on one day.