Advertised AI training pay: $3.30–$400/hr

A comparison of 1,177 USD-hourly listings with complete rate ranges. These are rates advertised in saved listings, not verified earnings, guaranteed offers or available paid hours.

Newest saved observation in this sample: . Individual records have different observation dates. See the sample and calculation method.

What the numbers tell you

  • Included advertised bounds span $3.30–$400/hr, in USD.
  • The median of listing-range midpoints is $60/hr. A midpoint is a comparison convention, not the rate offered to an applicant.
  • Qualifications come from the individual role. A pay band cannot tell you whether a degree, experience or particular location is required.
  • Compare the rate with the application conditions, payment unit and available work before deciding where to apply.

How advertised rates are distributed

Each listing goes into one band using the midpoint of its stated minimum and maximum. These editorial boundaries describe rates, not entry-level, specialist or licensed work. A listing's full range may cross a band boundary.

$30 or less

Advertised-range midpoint per hour

233 listings

19.8% of the included sample

Over $30 to $75

Advertised-range midpoint per hour

523 listings

44.4% of the included sample

Over $75 to $150

Advertised-range midpoint per hour

317 listings

26.9% of the included sample

Over $150

Advertised-range midpoint per hour

104 listings

8.8% of the included sample

Advertised hourly rates by category

Categories with at least three included listings appear below: 1,177 listings in total. These are listing-level comparisons, not a controlled comparison of applicants or employers.

CategoryMedian midpoint /hrAdvertised bounds /hrListings
Medical & Clinical$120$25–$35068
Legal & Compliance$102.50$9–$40043
Finance & Accounting$80$30–$30038
Software Engineering$75$15–$300215
STEM & Science$71.25$20–$400190
General / Generalist$60$4.50–$350132
Writing & Content$60$10–$35062
Output Evaluation$38$3.30–$20048
Voice & Audio$37.50$10–$280178
Data Annotation$34$3.50–$28575
Translation & Linguistics$30$4.20–$120128

For each field's 25th to 75th and 10th to 90th percentiles, with CSV and JSON downloads and a data dictionary, see the AI-training pay index.

What is in this sample?

We start with 3,107 distinct public listing records that pass the site's active, content and freshness checks. Those checks do not confirm that an employer is still hiring today. 23 other public records fail those checks and are left out before pay analysis.

1,177 records enter the numerical sample; 1,930 are excluded by the pay checks below. Counts are applied in this order, so each record appears in only one exclusion group.

  • 0 have another currency or no stated currency.
  • 1,347 remaining records have a non-hourly or missing pay unit.
  • 0 have invalid numerical bounds or a minimum greater than the maximum.
  • 583 have a missing endpoint or only an “up to” / “from” rate. Stored zero values mean missing pay, not unpaid work.
  • 0 fall outside the site's $3–$500/hr validation screen. These limits screen records; they are not the observed minimum and maximum.

We count each listing ID once; 0 repeated IDs were discarded. Different IDs may still describe related or repeated opportunities. We do not convert currencies, annual salaries or per-task amounts into hourly rates.

For each included record, the midpoint is (minimum + maximum) ÷ 2. The median is the middle midpoint, or the mean of the two middle values for an even-sized sample. The displayed range uses the lowest stated minimum and highest stated maximum. Calculations use unrounded values; amounts are displayed to at most two decimal places.

This sample reflects SignalGigs' saved inventory and coverage, not the entire AI job market. Listings carry equal weight, regardless of applicants, openings or hours. Regional and qualification differences remain mixed within categories. Records with incomplete pay are absent from the statistics, which can affect the result.

Saved observation dates in the included sample run from October 29, 2026 to November 11, 2026; 0 included records have no usable observation date. These dates record observations, not guaranteed current availability.

Compare offers before chasing the highest number

  1. Check your fit first. Read the role's degree, experience, subject, language and location conditions. Use the STEM listings or coding listings to compare roles in your field.
  2. Clarify the rate. Ask which part of an advertised range applies to your task and location, and whether assessments, preparation or revisions are paid. Do not treat the top of a range as your offer.
  3. Check the work and payment terms. Look for available hours, project duration, approval rules and payment timing. A high rate cannot establish the amount of paid work you will receive.
  4. Measure your own time. For a hypothetical $18 task taking 30 minutes, task time alone implies $36/hr. Add 30 minutes of unpaid preparation and that becomes $18 per hour of total time. Neither figure is a platform claim; use your actual payment and time records.

More ways to compare AI work

Each guide has its own scope and data date. Check its methodology before comparing figures.

Frequently asked questions

How much do AI training jobs advertise per hour?

The 1,177 included USD-hourly listings advertise bounds spanning $3.30–$400/hr. The median of their per-listing range midpoints is $60 per hour. These are advertised figures, not measured worker earnings or guaranteed offers.

How do evaluation and data annotation rates compare?

In this sample, evaluation has a median advertised-range midpoint of $38 per hour across 48 listings; annotation has $34 across 75. Differences in tasks, location and qualifications can affect the comparison. This does not measure what the same applicant would earn in either category.

Does a higher advertised rate mean a particular degree is required?

No qualification is inferred from the rate. Check the individual listing for degree, experience, language and location conditions. The rate bands on this page group numbers only; they do not classify applicants or establish eligibility.

Why are some listings missing from the pay tables?

The numerical sample requires an eligible public listing with USD currency, hourly pay and two positive, ordered rate bounds within the stated validation limits. Missing or one-sided rates, other currencies and other payment units are excluded rather than converted or filled in. Category rows need at least three included listings.

How should I compare hourly and per-task work?

Keep the payment units separate until you know the time involved. For a hypothetical task paying $18 and taking 30 minutes, task time alone implies $36 per hour. If another 30 minutes is unpaid preparation, the same $18 covers one hour. This is arithmetic, not an expected rate or a claim about any platform.

Can I turn a listed rate into an annual income estimate?

A listed rate does not establish available hours, accepted work, project duration or payment timing. Ask the hiring platform about those terms before planning around the income. This dataset contains advertised listing rates, not contractor payment records.

When were these listings checked?

The newest saved observation among included records is November 11, 2026. Records have different observation dates; this is not a check of every listing on that date. Open each listing for its own saved check information and verify availability with the hiring platform.

Find roles that match your background

Compare the full listing, its qualifications and its pay terms before applying.