Skip to main content
Skip to content
Agentic Finance

AI Microtask Pay in 2027: What You Earn After Unpaid Time, Rejections and Fees

Calculate AI microtask earnings after unpaid time, rejections and payout costs. A transparent 2027 worksheet separates pending rewards from spendable receipts.

Small task payment beside an hourglass illustrating paid and unpaid work time
Reading options: audio, text size & sections

Decentralised News · 2027 planning series · Sources reviewed 22 November 2026

What Matters

The useful pay rate for AI training and online microtasks is the money a worker receives after costs, divided by all the time required to earn it. Task prices, approved balances and spendable cash answer different questions. DN’s calculator follows one work cohort through unpaid preparation, rejected tasks, pending approval and payout so that an attractive advertised rate can be tested against a complete working day.

DN Evidence Block

2027 planning edition. Sources reviewed: 22 November 2026. This analysis combines current first-party payment and participation rules with an original accounting worksheet. Prolific and Outlier provide examples of different task-payment arrangements; they are not ranked or recommended. Amazon Mechanical Turk appears only as a documented closed service. The worked week is synthetic. DN has not completed paid tasks, audited worker accounts or measured representative 2027 earnings.

A $12 task-hour becomes a $6.11 working hour

Illustrative scenario: a worker completes twelve half-hour tasks. Each offers $6, making the visible task rate $12 an hour. The worker also spends two hours finding suitable work and completing qualifications, followed by one hour on corrections, an appeal and payment administration. The whole cohort consumes nine hours.

At the chosen payment cutoff, ten tasks have been paid: $60. One $6 task was rejected and another $6 remains pending. A hypothetical 3% payout and conversion charge, a further $0.70 fixed fee and $2.50 of allocated direct expenses reduce the receipts to $55. The realised receipt rate is $55 divided by nine hours, or $6.11 an hour.

None of those percentages describes a measured platform average. They are deliberately visible inputs. The example shows why the denominator changes the decision. If the pending reward later arrives, the cohort record improves. If a rejected task is successfully appealed, that record changes again. Neither amount should be quietly counted as cash before it is received.

There can still be reasons to accept this work: flexible timing, experience, a particular subject interest or a useful supplement to another income. The worksheet helps price that choice. It does not tell a worker what their time ought to be worth, determine employment status or imply that a better-paid alternative is immediately available.

The Signal: there are four different pay numbers

NumberWhat it measuresWhat it leaves unresolved
Advertised task rateOffered reward divided by stated task durationYour actual time, availability, acceptance and payout costs
Approved rewardWork accepted for payment under the applicable processWhen and how the funds become available to you
Gross receivedRewards actually paid for the selected work cohort by the cutoffFees, conversion and the expenses needed to earn them
Net receipts per total hourReceived money after measured costs divided by the complete work-time recordTaxes, benefits, future availability and the value of alternative uses of time

These figures should coexist in a useful earnings record. A task-rate metric helps compare offers before starting. An approval record identifies outstanding receivables. A receipts record supports the household cash plan. The net hourly figure tests whether the whole workflow is improving. Combining them into one number can hide the reason earnings changed.

1. Start the clock before the first paid task

Record three non-overlapping time categories: task execution, search and qualification, and rework or administration. Include the time spent on a task that is eventually rejected. Its commercial outcome changed; the time already spent did not disappear.

Historical research provides a reason to inspect the whole working process. The ILO’s 2018 digital labour platform study covered 3,500 workers across 75 countries and examined pay, work availability and working conditions on five microtask platforms. It is historical evidence about how to investigate platform work, not a current pay table for the AI training market.

Use a practical time boundary. A browser tab left open while you cook is not automatically an hour of active work. An hour spent monitoring a queue when you cannot use that time for anything else may matter to your decision. Maintain an active-work measure, then add a separately labelled availability measure if the distinction is material. Do not switch definitions between weeks simply because one produces a more attractive result.

One-time onboarding also deserves an explicit rule. You can charge it entirely to the first cohort or spread it across a stated number of completed tasks. Spreading it across hoped-for future work makes the present rate look better without establishing that those tasks will arrive. Keep the original onboarding total visible either way.

2. Treat the platform’s rate definition as a specific promise

Prolific’s participant guidance, dated 28 July 2026, sets a minimum study reward rate of £6 or $8 per hour and recommends £9 or $12 or more. It says its effective study rate uses median actual completion time. That is a documented study-payment policy. It does not promise nine paid hours whenever a participant has nine hours available.

A worker should therefore preserve the task’s offered amount, estimated duration, actual duration and eventual adjustment. If the displayed rate changes as completion data arrives, retain the relevant version. A later bonus belongs in the same cohort as the original task so that the before-and-after comparison remains meaningful.

Read the economic unit carefully when comparing research studies, annotation tasks, expert review and project-based AI evaluation. An hourly project rate, a fixed amount per accepted task and a reward tied to an estimated completion time are different arrangements. Normalize them only after identifying the time and acceptance conditions each actually covers.

Outlier’s terms effective 11 May 2026 link task payment to completion and verification, and describe circumstances in which payment can be delayed or suspended during a compliance investigation. Its contractor language should be read as that provider’s terms, not as a legal conclusion about every worker in every jurisdiction. For the worksheet, the immediate implication is narrower: submission and confirmed payment remain separate events.

3. Keep rejected, pending and approved-unpaid work separate

A useful ledger gives each task a stable identifier and a current state. Suggested states are submitted, approved, paid, rejected, appealed and cancelled. Preserve the timestamp and evidence when the state changes. A rejection under appeal should retain both the rejection and the appeal record rather than becoming a paid task by assumption.

Prolific’s cash-out instructions, dated 6 August 2026, distinguish pending rewards from available balances. They explain that rejected rewards do not become available and that approved rewards are paid through PayPal. These mechanics illustrate why a dashboard total can contain money at different stages.

In the synthetic week, the $6 pending reward is neither written off nor spent. It sits in a separate column. If it is paid later, update the receipt cutoff and keep the original record. That allows two useful questions: what was spendable when a bill fell due, and what did the cohort eventually earn?

Appeals need their own economics without becoming a reason to surrender a valid claim. Record the amount disputed, time already spent, available evidence and next permitted step. A worker may reasonably pursue a small dispute to correct a record or clarify a policy. The calculator simply ensures the time is visible instead of making successful collection appear effortless.

4. Follow payment into the currency you actually use

Choose one reporting currency for the entire cohort. Preserve the original-currency reward, payout date, amount received after conversion and separately disclosed charges. If you use an effective percentage for convenience, state how it was calculated. A rate inferred from one payout is an observation about that payout, not a promise about the next one.

Do not deduct the same conversion loss twice. If the receipt entered is already net of every fee, set the corresponding fee inputs to zero. If it is the gross platform reward, subtract the actual or explicitly assumed charges once. Fixed charges and proportional charges should remain separate because batching payments can affect them differently.

For cross-border work, the final usable balance may be several steps away from the task platform. The person comparing work should establish the available payout method in their own country and account, the receiving currency and any onward conversion or bank-transfer requirement. A provider’s supported payment logo alone does not answer those questions.

DN’s freelancer payment-rail research addresses receiving, spending and cashing out stablecoins. This article has a different decision boundary: whether the paid task is attractive after all work and payment costs. A new payment rail cannot repair an uneconomic task merely by settling it faster.

5. Give the work cohort a receipt cutoff

Calendar totals can create misleading comparisons. A quiet week may contain a large payout for last month’s work, while an intensive week may finish before approval. Dividing each week’s deposits by that same week’s hours makes one look exceptional and the other disastrous even when task economics did not change.

The DN method follows a matched cohort: the same identified tasks, the time required for those tasks, their allocated costs and the receipts received by a declared date. Report an initial cutoff and, where useful, a final settled view. This keeps payment delay visible without mistaking delayed receipts for permanently lost earnings.

There is a related limit on comparisons between workers. Different eligibility, expertise, equipment, queue access and available hours can produce different task mixes. One person’s screenshot is not a representative wage survey. A credible public comparison would need a disclosed sampling method, a consistent time definition and treatment of people who received no suitable work.

6. Automation must respect the task’s rules

An article about AI work should not imply that every task can be delegated to another AI. Prolific’s AI-assistance guidance says participants must not use AI assistance for submissions unless the researcher specifically requests it. The applicable instruction for the particular study or project matters.

A compliant personal workflow can still make recordkeeping easier: a timer, an expense sheet and a reminder to check a payment status. Keep confidential task material, participant information and assessment content out of unrelated tools unless the relevant rules permit that use. Faster output is not an economic gain if the method invalidates payment or exposes protected material.

For a worker building a broader client service, DN’s Agentic Freelancer Stack covers lead research, delivery and client operations. Platform microtasks should be evaluated on their own terms before being treated as a stepping stone to that separate business model.

7. Check whether the earning opportunity still exists

Older earning lists can preserve a familiar brand long after the operational decision has changed. On the source-review date, Amazon Mechanical Turk’s official closure FAQ states that the service permanently closed on 30 September 2026. Its worker guidance addresses remaining payments and gives 30 November 2026 as the payment-preferences deadline. It is therefore excluded from earning-platform recommendations here.

The wider lesson is a status check before investing time in an opportunity. Look for an accessible official application route, current eligibility, active work terms and a documented payout process. An archived tutorial, old referral page or undated earnings claim does not establish that new work is available to you.

DN Microtask Receipt-Rate Calculator

Enter aggregate figures only. No task content, identity document, account credential or payment-account connection is needed. The calculator reports net receipts, total work hours and receipts per hour. Approved-but-unpaid and unapproved amounts remain visible outside the cash calculation.

DN Microtask Receipt-Rate Calculator

Follow one matched work cohort through total time, payout costs and a stated payment cutoff. Use aggregate figures in one currency; do not enter private task or account information.

This instrument runs locally in your browser. It does not connect to accounts, fetch evidence or submit instructions. Use aggregate values and non-sensitive references. Downloaded records contain your inputs.

Enter your inputs or load the labelled example, then select Calculate.
Machine-readable illustrative record

An illustrative record of the example assumptions, not an observed transaction or provider test. A null calculation timestamp identifies this static example.

{
  "schema_version": "1.0.0",
  "article_id": "N003",
  "evidence_basis": "illustrative",
  "source_review_date": "2026-11-22",
  "computed_at": null,
  "inputs": {
    "task_minutes": "360",
    "search_minutes": "120",
    "admin_minutes": "60",
    "gross_received": "60",
    "variable_fee_pct": "3",
    "fixed_fees": "0.7",
    "direct_costs": "2.5",
    "approved_unpaid": "0",
    "pending": "6"
  },
  "status": "Receipt analysis calculated",
  "outputs": {
    "hours": 9,
    "net": 54.99999999999999,
    "effective_rate": 6.111111111111111,
    "gross_task_rate": 10,
    "approved_unpaid_excluded": 0,
    "pending_excluded": 6,
    "reporting_basis": "matched_work_cohort_receipt_cutoff",
    "evidence_verified": false
  }
}

How to make the next week’s decision

Compare two complete cohorts before changing your workflow. If the task rate is stable but the realised rate falls, inspect queue search, qualification time, rejection, payment charges and administrative work separately. Each points to a different response. Changing the payout method will not resolve slow task selection; completing tasks faster will not resolve a missing approval.

Set a bounded experiment instead of extrapolating a strong day into a monthly salary. Define how much time you can allocate, what costs you will incur, which task categories you are eligible to attempt and when you will review the result. Retain unsuccessful attempts in that review. A useful result may be to continue, narrow the task type, request clarification or stop allocating additional time.

Keep household commitments tied to money available when needed. Pending work may ultimately pay, but a bill has its own deadline. This is why the same cohort needs both an eventual-earnings view and a payment-cutoff view. The latter measures a timing constraint that an impressive final hourly rate can obscure.

DN Alpha Thesis

The worker’s advantage begins with a complete earnings record. AI task markets become easier to evaluate when offered rewards, acceptance, time and spendable receipts remain separately visible. A task that looks ordinary can prove worthwhile with little unpaid overhead. A prestigious headline rate can weaken after repeated qualification, uncertain approval and expensive payout. Measure the whole route before scaling the commitment.

Methodology and limitations

DN’s original method defines total hours as task, search/qualification and administration minutes divided by sixty. Net receipts equal gross paid rewards less the selected percentage charge, fixed charges and allocated direct expenses. Net receipts divided by total hours produce the realised receipt rate. Negative net results remain valid; zero total time cannot produce a meaningful hourly rate.

The example uses twelve hypothetical tasks, not a tested platform. Provider documents establish the specific policies cited; they do not verify a particular worker’s eligibility, task availability, receipt timing or earnings. No 2027 market-wide performance is claimed. Taxes, statutory benefits, employment classification, depreciation and the value of foregone alternatives require separate treatment where relevant. Update the source review whenever payment terms or operating status change.

Frequently asked questions

Is the advertised task rate my hourly wage?

It measures the offered task economics under a stated duration. Your complete working-hour result also includes search, qualification, rework, acceptance and payment costs. Preserve both figures so their difference can be explained.

Should pending rewards count as income in this calculator?

They remain a separate amount. The main result measures receipts paid by your selected cutoff. Update the same cohort if the pending money is later approved and paid.

Do I include time spent on rejected tasks?

Yes. The time belongs to the work cohort even when payment is rejected. Also record the appeal outcome and the time spent resolving it.

Can I use AI to complete paid studies?

Only when the specific study or project rules allow it. Prolific’s cited guidance generally prohibits AI assistance unless the researcher explicitly requests it. Other services have their own terms.

Why can my net result be negative?

Your measured direct expenses and payment charges can exceed the rewards received by the cutoff. The calculator keeps that result visible rather than replacing it with zero.

Does this article recommend a platform or predict my earnings?

No. It provides a transparent method using dated policy examples and a synthetic worked week. It does not establish available work, admission, legal status or a future earnings rate.

Research and disclosure: DN can apply this method to an agreed, anonymized task-payment sample and produce a documented earnings analysis. That service does not secure work or guarantee income. No task platform paid for inclusion, and no platform-referral recommendation is made in this article.

Primary sources and review record

  1. ILO: Digital labour platforms and the future of work. Historical 2018 study covers 3,500 workers in 75 countries and investigates pay, work availability and working conditions; not a 2027 wage estimate. Reviewed 2026-11-22.
  2. Prolific: Underpaying studies. 28 July 2026 policy gives minimum £6/$8 per study-hour, recommended £9/$12 and median actual-completion-time basis. Reviewed 2026-11-22.
  3. Prolific: How do I cash out?. 6 August 2026 documentation separates pending and available rewards and describes PayPal cash-out; rejection does not create available funds. Reviewed 2026-11-22.
  4. Outlier: Terms of Use. 11 May 2026 terms tie payment to completed verified tasks and describe investigation-related payment delay/suspension; no universal employment-status conclusion. Reviewed 2026-11-22.
  5. Prolific: AI assistance in submissions. AI assistance generally prohibited unless researcher explicitly requests it. Reviewed 2026-11-22.
  6. Amazon Mechanical Turk: Closure FAQs. Official page states permanent closure on 30 September 2026 and remaining payment-preferences deadline of 30 November 2026; service excluded from earning recommendations. Reviewed 2026-11-22.

Author and original DN instrument: Heath Muchena. Edition: 2027 planning. Source review: 22 November 2026. Version: 1.0.0. First publication adds the sourced article, declared-input tool and record schema. Refresh when the cited specification, provider terms, case status or source methodology changes.

Corrections: press@decentralised.news. No independent audit or transaction test is claimed. No paid placement or affiliate recommendation is included in this article.

Get the most talked about stories directly in your inbox

Join the Decentralised News briefing for independent crypto, DeFi and AI analysis. No spam, unsubscribe anytime.