Every "how much does this pay?" question hides a second question that matters more: how is the pay computed? Two platforms can advertise similar-sounding work and produce completely different earnings — not because one is a scam, but because one pays per completed unit and the other pays by the hour you can bill. Across the 245 platforms we track, 14 distinct pay models are in play, and knowing which one runs behind a platform tells you what your month will actually look like before you sign up.
The four models that cover most of the market
Points systems are the most common single model — 77 platforms, and 75 of them are survey panels. Your dashboard shows a balance in points; the real numbers hide in the conversion rate. Swagbucks points convert to roughly $0.50–$2.50 per survey, but a panel whose points are worth a tenth of a cent can make a 500-point survey look impressive and pay fifty cents. Always find the conversion page before judging a points balance.
Per-task pay dominates micro-work and AI-training platforms: 71 platforms pay this way, 31 of them in AI training and data labeling. Each accepted task pays a fixed amount — a few cents to a few dollars each at Clickworker, for example. The model is transparent (you can see the price before you start) but unforgiving: rejected work pays nothing, and your effective rate collapses whenever a queue is empty.
Hourly or project rates are the closest thing to a job: 44 platforms, concentrated in AI training (29 of them). Outlier, for instance, runs general projects at roughly $15–40/hr with specialist domains higher. The catch is access — hourly platforms typically gate entry behind assessments, and project availability still fluctuates. But when work flows, this is the only model where a platform's advertised number can meaningfully beat local part-time wages.
Per-study or per-session pay covers 41 platforms — 31 in surveys (focus groups, in-depth interviews) and 8 in app and website testing. This is the highest-variance model: a 90-minute focus group can pay more than a week of microtasks, but the studies are infrequent and often screen hard. Treat per-study platforms as opportunistic income, never as a base.
The specialists: audio minutes, messages, and devices
Transcription runs on its own clock: 15 of the 16 per-audio-minute platforms are transcription platforms. GoTranscript advertises up to $0.60 per audio minute — $36 per hour of audio — but the real-world rate depends on how long that audio takes you to transcribe. A difficult file at 4:1 effort-to-audio ratio pays a quarter of the headline number. Per-audio-minute is the model where advertised and effective rates diverge the most.
Online chat work is message-metered: four platforms (Phrendly, TexKings, Text121Chat, Texting Factory) pay per message — around $0.35 per message is commonly cited — sometimes blended with per-minute rates. And the rarest model is genuinely passive: four platforms (MobileXpression, SavvyConnect, UpVoice, and MSR) pay a small monthly amount simply for keeping an app or extension installed that shares browsing data. Nobody gets rich at it — think a few dollars a month — but it is the only model with zero active work.
Hybrids, and why the model predicts the experience
43 platforms run more than one model at once — usually a points core with a cash-per-referral or per-study side. When you compare platforms, match models first: an hourly AI-training project and a points survey panel are not competing for the same hour of your day. And when a platform's rate looks too good to be true, check whether the pay model can even deliver it mathematically — a per-task platform that promises full-time income is promising you an endless task queue, which is the part they never advertise. Every platform page in this directory lists its pay models, verified, so you can check the shape of the work before you hand over an email address.