A remote AI training job advertises $50 an hour. You get selected, complete the work, and earn $1,000. Sounds like 20 hours of work, right?
Not necessarily.
If you also spent ten hours on project preparation, reading instructions, revisions, and administration, you devoted 30 hours to earning that $1,000. Your effective hourly earnings were $33.33, before any withdrawal fees or taxes.
That doesn’t mean the advertised $50 rate was false. It means the rate answered a different question: How much is an approved, payable hour worth? Not how much will I make for every hour I put into this opportunity?
If you’re considering remote AI training work as a side income or a serious freelance career, this distinction matters more than an impressive number in a job listing.
Table of Contents
The two hourly rates every AI contractor should know
Advertised hourly rate is the amount offered for an hour of work that qualifies for payment under a project’s rules.
Effective hourly rate measures what you actually earn for the time you spent securing, preparing for, and completing the work.
Here’s the basic calculation:
Effective hourly earnings = Total earnings ÷ Total time invested
If you want the amount left after transaction costs, use:
Effective hourly earnings after fees = (Earnings − applicable fees) ÷ Total time invested
This is a personal planning measure, not a replacement for the platform’s official hourly rate. It’s particularly useful when comparing opportunities with different workloads and onboarding requirements.
Follow the hours, not just the dollar signs
Suppose you join a project advertising $50 per paid hour. Here’s how one illustrative engagement might unfold:
| Activity | Hours | Paid? |
|---|---|---|
| Initial onboarding and setup | 4 | No, in this example |
| Reading guidelines and preparing | 2 | No, in this example |
| Approved project work | 20 | Yes |
| Corrections and revisions | 3 | No, in this example |
| Reporting and payment administration | 1 | No |
| Total | 30 | 20 paid hours |
The project generates 20 × $50 = $1,000 in gross earnings. But it takes 30 hours of your time.
Your effective rate is $1,000 ÷ 30 = $33.33 an hour. Only two-thirds of the time you invested generated payment in this scenario.
Crucially, these unpaid activities are assumptions for the example, not universal industry rules. Some projects pay for onboarding or revisions; others have different conditions. Read the actual agreement before accepting work.
Also distinguish active work from simply waiting for a project to become available. You wouldn’t normally count a week in which you did nothing as 168 hours of work. But time spent actively interviewing, taking assessments, or preparing applications is worth tracking.
Three different kinds of time can appear on your dashboard
It’s easy to assume that time logged by a platform is automatically time you’ll be paid for. That’s not always how it works.
For example, Terac explains that its hourly projects distinguish tracked time from payable time. According to its project documentation, a project may pay for all tracked time, productive time only, or productive and neutral time, depending on its rules.
Similarly, a Mercor expert-pool listing explains that candidates can complete screening and join a pool before receiving an invitation to a specific project. Being in the pool isn’t itself a paid engagement.
Think in three categories:
- Time invested: Everything you actively do to pursue and deliver the opportunity.
- Time tracked or submitted: Hours recorded in a timer or timesheet, if the project uses one.
- Time approved for payment: Hours or tasks accepted under the contract’s payment rules.
These numbers can differ. The important question is not just whether you completed the work, but what the client considers payable.
The fixed-fee trap: a $12 task isn’t always a $36-per-hour job
Hourly jobs aren’t the only way AI training work is compensated. Some assignments pay a fixed amount per response, recording, interview, or completed task.
Imagine a project offering $12 for each approved task.
If a task takes 20 minutes, you can theoretically complete three per hour. That’s $36 an hour, assuming the tasks are available, all are approved, and you can sustain that pace.
But if each task takes 35 minutes, the same $12 reward works out to roughly $20.57 an hour.
Add time spent studying new guidelines or fixing submissions, and your effective rate could be lower still.
Before accepting per-task work, estimate a realistic completion time. Don’t calculate pay using only the fastest task you’ve ever finished.
The hidden costs don’t all come from unpaid work
Even when every hour is properly approved, the money deposited into your bank may differ from the amount shown in your dashboard.
Potential deductions include payout-provider fees, currency-conversion costs, bank charges, and taxes applicable to your circumstances. Not every payment method has every fee, and the amounts vary.
For instance, Terac’s payment documentation says that moving approved earnings from its balance to its payout balance is free, while the payout provider may charge fees for withdrawing to a bank. It displays the applicable charge before confirmation.
Return to our $1,000 example. If the combined transaction costs were hypothetically $20, you’d retain $980 before tax, giving you an effective rate of $32.67 per hour over those 30 hours.
That $20 is an illustrative deduction, not Terac’s fee or a claim about any other platform.
If you’re paid in US dollars but spend in another currency, compare the actual amount your bank receives with the gross amount due. A favorable-looking advertised rate doesn’t tell you the exchange rate or charges used in the final transfer.
Higher hourly pay doesn’t always mean better monthly income
Let’s compare two hypothetical offers.
| Project A | Project B | |
|---|---|---|
| Advertised rate | $70/hr | $35/hr |
| Payable hours available | 8 | 20 |
| Gross earnings | $560 | $700 |
| Additional active unpaid time | 3 hours | 2 hours |
| Total time invested | 11 hours | 22 hours |
| Effective hourly rate | $50.91 | $31.82 |
Project A pays more for each hour of your total time. Project B generates more total money.
Neither is automatically the better choice. Someone with limited free time might prefer Project A. Someone who wants a larger monthly payout and has more availability might choose Project B.
The sensible comparison depends on your goal: maximizing income, maximizing earnings per hour, developing specialist experience, or securing relatively predictable work.
The same problem applies to very high-paying specialist listings. A $150-per-hour project that supplies only a handful of paid hours may generate less income than a lower-rate project with an ongoing workload.
Our separate guide on how much AI training jobs actually pay covers the broader differences between generalist and specialist compensation. Here, the point is how to judge the economics of a specific opportunity.
Approved work matters more than completed work
Some AI projects review submissions before confirming payment. That’s not inherently unusual, but it changes what you need to track.
Terac, for example, says one-time work is credited when approved, while hourly and per-task projects follow their own payout schedules. Its reward policy also explains that applicants who are screened out of a study aren’t paid for that screening.
Some micro1 data-collection opportunities likewise advertise pay for approved recordings, with project-specific guidelines. That does not establish that every micro1 project uses identical payment conditions.
It’s worth separating submitted earnings, approved earnings, and money actually received in your records. If there’s a dispute, that trail is more helpful than remembering only the headline rate.
Don’t automatically assume every revision is unpaid, either. Ask whether corrections count as paid time and whether rejected work can be appealed. A strong opportunity makes these expectations clear.
Your first week and your fifth week may look very different
One-time onboarding costs can make a new contract look less attractive in its first week than it does later.
Suppose you spend four unpaid hours learning a platform and then earn $500 for ten payable hours in the first week. Your effective rate for that week is $500 ÷ 14 = $35.71.
If the same project continues for four more weeks, producing $500 for ten hours each week without additional onboarding, your cumulative earnings become $2,500 over 54 hours: an effective $46.30 per hour.
That’s why you should distinguish one-time setup costs from recurring unpaid work.
But don’t assume a short contract will run for months. An opportunity that ends after the first week never gets the benefit of spreading those initial hours over future earnings.
A simple earnings tracker makes the difference visible
You don’t need sophisticated financial software. A spreadsheet with these columns is enough:
| Column | What to record |
|---|---|
| Platform and project | Which opportunity generated the work |
| Advertised rate | Hourly amount or per-task reward |
| Active unpaid time | Applications, onboarding, prep, unpaid revisions |
| Payable hours or approved tasks | Work that qualifies for payment |
| Gross approved earnings | Amount earned before deductions |
| Payment charges | Withdrawal, conversion, or bank fees |
| Net received before tax | Amount actually received after payment charges |
| Effective hourly rate | Net received before tax ÷ all active hours invested |
Track projects individually, then look at monthly totals. You’ll be able to tell whether a high-rate contract actually outperforms a lower-rate one after accounting for preparation and administration.
When a payment hasn’t arrived yet, label it approved but unpaid rather than treating it as cash received.
Five questions to ask before accepting the next offer
Before you commit to a project, try to get clear answers to five things:
- Exactly what earns payment? Approved hours, tracked hours, completed tasks, or a fixed deliverable?
- Which steps are unpaid? Assessments, onboarding, guideline review, or revisions?
- How much paid work is realistically available? A few tasks, a capped weekly schedule, or an ongoing queue?
- When does the money become withdrawable? After submission, review, approval, or a fixed payout date?
- What will reach your account? Which fees, currencies, and withdrawal methods apply?
A higher hourly rate is still worth pursuing when the work suits your expertise. Just don’t let that single number make the entire decision.
The real measure of an AI training opportunity is what it pays for your time, not merely what it advertises per payable hour. Once you start tracking that distinction, comparing projects becomes much easier.
Disclosure: This article contains referral links. I may earn a commission when eligible applicants complete qualifying activities through certain platforms, at no additional cost to them. Referral links do not guarantee selection, ongoing work, or earnings.
Leave a Reply