People paid between $45,000 and $75,000 each, plus another $10,000 or more for stock, to buy into a scheme promising passive income from a “proprietary system powered by artificial intelligence”. It claimed affiliation with Nike and Disney. It sold on Amazon, Walmart and TikTok.
The Federal Trade Commission took it to court, and the complaint contains the company’s own figures. More than a fifth of the stores it set up for customers earned nothing at all. Another third earned under $2,500 in gross lifetime sales.
That is the market you are walking into when you search for ways to make money with AI. It is the most scam-saturated corner of the internet, and it is saturated precisely because the underlying premise is half true: AI did create new ways to earn, some of them accessible and that truth is what makes the lie sellable.
So this section does the boring thing nobody else does. We went looking for every AI earning route we could find, checked which ones have evidence of real money behind them, and put the dead ends and the frauds on the same page as the working routes rather than in a separate article nobody reads.
Set your expectations from real data first#
Bankrate’s 2026 survey found about 27% of American adults have a side hustle, down from a 36% peak in 2024. The median earner makes $200 a month. The average is $885, dragged up by a small number of people doing well. In the previous year’s survey, 28% of side hustlers were earning between $1 and $50 a month.
That is the honest baseline for anything in this section. Hundreds of dollars a month is a realistic good outcome. Thousands is possible on two of the routes below if you have the right background. “Replace your salary in 90 days” is the sales pitch of the thing the FTC prosecuted.
The scale of each route#
| Route | Evidence it pays | Works outside the US? | Realistic money |
|---|---|---|---|
| Working for the AI labs | Strong. Primary reporting, named figures | Partly, the good rates are US and Anglosphere | $40–110/hr where you qualify; far less on the worldwide platforms |
| Automations for small businesses | Good. Marketplace rates plus demand surveys | Yes | $40–100/hr, projects $1,800–4,500 |
| Ordinary services, done faster with AI | Moderate. The margin logic is sound | Yes | Your normal service rate, at better margin |
| Making video with AI | Good on the demand side. IAB buyer surveys, platform policy in writing | Yes | Your production rate. Client work, not a channel |
| Building software without coding | Weak for beginners. No verified case of a true novice earning a living | Yes | Unmeasured. The base case is zero and a security liability |
| Selling AI images, prompts, GPTs | Banned or collapsed. Verified from platform policy | — | Not a route |
Two things about that table are worth saying out loud, because they are the opposite of what the genre usually claims.
The best-evidenced AI earning route is a job. Not a passive income stream, not an automated business. Domain experts get paid by the hour to help train models, and the money is real and documented. It is also the route with the least mystique and therefore almost no coverage.
The routes marketed as effortless are the ones with no evidence. Sell AI art. Sell prompts. Build an app in a weekend. Every one of those is either banned by the platform, competed into nothing, or supported only by success stories that fall apart when you check who the person was before they started.
What changed, and what didn’t#
The genuine shift is this: AI created enormous demand for two things that ordinary people can supply. Expert human judgement, because the labs cannot improve models without people who actually know medicine or law or senior-level code. And implementation, because 89% of small businesses now use AI in some form while 73% of them say they lack the training and resources to do it properly. Both of those are people-shaped problems with budgets attached.
What did not change is the part everyone wants to be true. AI did not make it possible to earn without doing anything, and it did not remove the need for a skill or a credential or a client. Where it removed the need for those, it also removed the price: that is the entire story of what happened to stock imagery, and it is happening to UGC rates now.
The pattern across every route below, without exception: where AI makes something easy, that thing stops being worth money. What pays is the part AI made scarce. Expert judgement got scarcer. Implementation capacity got scarcer. Generic output got worthless.
How to spot the scam version#
You will meet the fraudulent variant of this before you meet the real one, because the fraudulent variant has an advertising budget. It is recognisable:
- An upfront fee. The FTC case above ran on $45,000 to $75,000 entry payments. Legitimate work pays you. Nobody charges you to do data annotation, and no genuine client asks for a setup fee from a freelancer.
- “Done for you” and “passive”. Automated storefronts, hands-off AI businesses, systems that run themselves. This is the specific shape the FTC prosecuted, twice, it also ran an enforcement sweep against deceptive AI claims in September 2024.
- Income screenshots with no denominator. One person’s $12,000 month tells you nothing without knowing how many people tried. In the Click Profit filings, the denominator was that a fifth earned nothing.
- Borrowed credibility. Claimed partnerships with recognisable brands. Click Profit invoked Nike and Disney.
- Urgency and a closing cohort. A real opportunity that pays by the hour does not need you to decide tonight.
- The course is the product. If someone is selling you the method rather than doing the thing, the method is the business and you are the market.
Bankrate 2026 for side-hustle earnings. The small-business percentages are reported second-hand. Sources: FTC press release on Click Profit, case file, CNBC’s report, September 2024 crackdown on deceptive AI claims.
Where to start#
If you have a professional credential or deep domain skill — medicine, law, finance, senior engineering, professional writing, a trade, start with working for the AI labs. It pays the most, it pays by the hour, and your existing expertise is the qualification. Check the country list before you invest time.
If you are practical and like fixing things — automations for small businesses. The demand is documented, the tools cost tens of dollars a month, and the clients are local businesses who know they are behind.
If you already sell a service — read this before buying more tools. The opportunity is margin, not new customers.
If you already do video work — making video with AI. The demand is on the advertising side, the platform rules are narrower than most people assume, and the trap is pricing anywhere near what the render costs. The craft is split by length: short-form, long-form, and the quality bar under both.
If you were about to spend a weekend building an app to sell — read the reality check first. Not to talk you out of building, but because the evidence about earning from it is much thinner than the marketing, and the security data is alarming.
If you were planning to sell AI images, prompts or custom GPTs — save yourself the time.
And if none of these fit, the eight non-AI routes have not gone anywhere. Two of them still pay faster than anything on this page.
This is the fastest-changing section of the site.