What the labels mean

How much of the business you own, from paid tasks to owning a company.

Every entry sits on a five-step ladder. 1 Paid tasks: a platform finds the work and sets the pay. 2 Sell your skill: you set your price and win clients. 3 Run a service: you own the client relationship. 4 Build an asset: a product or audience that earns without your hours. 5 Own a business: capital, staff or premises. Lower steps pay sooner; higher steps take longer but have no ceiling.

Something that pays this month, paired with something you could own later.

The quiz pairs an option on ladder steps 1–2 (money soon) with one on steps 3–5 (something you own), linked by skills they share or a natural next step.

How well-sourced a page or fact is. Not a rating of the company.

A: confirmed by an official or primary source. B: reliable press, research or a dataset with a method. C: self-reported by workers or founders, or a single aggregator. D: unconfirmed, vendor or marketing content, or not checked yet. A low grade means we know less, not that the option is bad.

Official, press & research, dataset, community, vendor.

Official: the company's own docs, pricing or terms. Press & research: journalists, researchers, legal analysis. Dataset: aggregators that state their method. Community: workers, founders, podcasts (self-reported). Vendor: marketing or sponsored content, which is the weakest.

Who says it: the company, the press, workers, marketing, or us.

From the company: stated by the platform itself. Press & research: reported by journalists, researchers or aggregators. Workers & founders say: self-reported experience. Marketing claim: not independently checked. Our estimate: calculated by us from the other facts.

Checked: we looked again and it was still right. Updated: something changed.

'Checked' is the last time the facts were re-read against their sources. 'Updated' is the last time the page's content actually changed. Stamps fade as a check gets older and turn into 'Re-check due' when it's overdue.

What people typically get vs. the best rate a platform advertises.

Advertised pay is the range or maximum a platform promotes. Realistic pay is what listings and workers report for most people, with dates. The bar on each page shows both.

How many days from signing up until your first payment, as reported.

A range from the fastest to a typical case: application, tests, onboarding, first task and the platform's payout schedule all add up.

Money you need before you earn anything.

Tools, inventory, fees or equipment you pay for up front. Most platforms cost nothing to join; business models usually cost something.

Open, turbulent, waitlist, pilot, unverified, paused or closed.

Open: accepting people. Turbulent: open, but with recent bans, cuts or lawsuits. Waitlist: sign-ups are queued. Pilot: a limited test. Unverified: not checked yet. Paused: temporarily not accepting. Closed: no longer accepting new people.

A referral link that may pay us a fee if you sign up. Always labelled.

Paid links are marked 'Ad · paid link' where they appear. They never affect facts, grades or rankings. See How we make money.

A founder's own account of their business, usually from an interview.

Revenue figures are what founders say about themselves and aren't verified unless marked ('showed a dashboard' or 'verified via payment provider'). They show what's possible, not what's typical.

How many founders reported success with this model in the last 12 months.

Strong: five or more case studies in the last 12 months. Some: two to four. Thin: one. None: no recent case we know of. More evidence means more people are succeeding with it right now, not that it's easy.

Tests or onboarding you must do before you get paid work.

Many AI-training platforms ask for assessments, interviews or onboarding that aren't paid. Some allow only one attempt.

An app that records your screen or activity while you work.

Some hourly platforms require a desktop app that logs activity or screenshots to verify hours.

Paid tasks that teach and test AI models.

Writing ideal answers, ranking model responses, checking facts or code, and creating hard test questions, often for AI labs through platforms like Mercor or Outlier. Experts in fields like medicine, law or coding usually earn the most.

Monthly recurring revenue: subscription income per month.

Used for software and memberships. $10K MRR means about $10,000 a month from subscriptions, before costs.

A small software product, often run by one person, sold by subscription.

Usually solves one narrow problem for a specific group. Costs little to start with AI coding tools, but most earn little; distribution is the hard part.