How to Make Money with AI

The truth about making money with AI

There is an enormous amount of "get rich with AI" material online. It does not work that way. AI accelerates real work; it does not print money. The paths below are real, and every one of them requires skill, effort and consistency — the advantage goes to people who already had something to offer and now deliver it faster.

What actually changed, and what it did to prices

This is the part the opportunity content leaves out, and it determines which of these paths is worth your year.

AI collapsed the cost of production. Writing, design, editing, voice, code scaffolding — things that used to take hours and a trained person now take minutes and no training. That is a genuine shift and it is why there is opportunity here at all.

But it collapsed that cost for everyone at once, including your competitors and including your potential clients. Anything whose value came from being laborious to produce is now worth much less, and falling. Writing a thousand words is no longer scarce. Making an image is not scarce. Assembling a video is not scarce.

What did not get cheaper, and is therefore where the money now sits:

• Distribution. Having anyone's attention at all. Unchanged in difficulty, and more valuable than ever because supply exploded.
• Trust. Being the person a business will let near its systems, its customers or its money.
• Knowing which problem is worth solving. Domain knowledge. A model can build the automation; it cannot tell you that invoicing is the bottleneck in a plumbing business.
• Accountability. Someone who owns the outcome when it breaks. This is most of what clients are actually buying.

The test to apply to any idea

Before committing months to anything on this page, ask one question: is AI the product, or is AI how I deliver something people already pay for?

If AI is the product — an app that summarises things, a service that writes posts, a tool that generates images — you are competing with everyone who had the same idea this week, plus the model vendors who may ship your feature for free. That market's price goes one direction.

If AI is how you deliver something with existing demand — a business that already pays someone to do this, badly and slowly — you have a business with a moat made of the thing you know and they do not. The buyer is not paying for AI. They are paying for the outcome, and they would not care if you used a spreadsheet.

Almost every durable success in this space is the second kind, and almost every piece of opportunity content describes the first.

How long each of these actually takes

No numbers here, because anyone quoting you a figure is guessing or selling. The shape is knowable and it is what people get wrong:

• Services — the fastest path to a first payment, because someone with a problem can pay you this month. Income tracks hours worked from the start, which is both its virtue and its ceiling.
• Automation for businesses — slower to the first client, higher value per client, and the work compounds as you reuse what you have built.
• Digital products — nothing for a long time, then small amounts repeatedly, if the audience exists first. Building the product before the audience is the standard way to earn nothing.
• Content and platform revenue — the slowest of all, measured in many months of unpaid work before anything meaningful, and the majority stop during it.
• Affiliate — follows content, so inherits its timeline.

The practical consequence: if you need income soon, start with services, and fund the slower assets with the proceeds. Starting with the slow one and hoping is the most common way this goes wrong.

Freelance services, and the trap in them

The fastest route: offer a service and use AI to deliver it faster — content, design, video and audio editing, translation, research. Sell on freelance platforms, in communities, or directly.

The trap is specific. If you compete on producing faster, you are competing on a cost that keeps falling for everyone, and the price of the deliverable falls with it. Charging by the hour actively punishes you: the better your tooling, the fewer hours you bill.

What works instead:

• Price the outcome, not the time. "Twelve posts a month, on brand, scheduled" is a price. "Six hours of writing" is a race to the bottom you will win and regret.
• Sell the judgement part. Anyone can generate copy; knowing what to say to this audience is the service.
• Own a niche. "Content for accountancy practices" beats "content" — narrower, easier to be found for, and you can charge more because you understand the subject.
• Move up to retainers. One client paying monthly is worth several one-off jobs, and the second month costs you less than the first.

Automation for businesses — the strongest path

Building automations and assistants that save a business real hours, and charging for it, is the most reliable route on this page. Demand exceeds supply, buyers have budgets, and the value is measurable in a way content never is.

Why it works: the client is not buying software, they are buying the end of a recurring annoyance. That is easy to price, easy to justify internally, and hard for them to do themselves — not because the tools are difficult, but because nobody there has the time or the inclination to learn them.

Finding the first client is the whole hurdle, and the answer is almost always proximity. A business you already know, or one in an industry you have worked in, where you can name the process that wastes time. Cold outreach describing "AI solutions" performs badly; a message describing their specific problem does not.

• Start with one process, end to end, and make it work properly — including the failure handling. A reliable small automation buys the second project.
• Quote on the saving, not the build. If it returns a day a month to someone, that is the frame for the price.
• Charge for maintenance. Integrations break, APIs change, the business changes. A monthly fee is honest and it is where this becomes a business rather than a series of jobs.
• Hand over documentation and access. Clients who cannot see inside what you built are clients who will not give you the next project.

Content, and the policy nobody mentions

An audience that earns from advertising, sponsorship and affiliates: faceless channels, a YouTube pipeline, a blog that ranks and keeps earning.

Two honest cautions before anyone builds a content business on generated material. Platform monetisation programmes require content that is meaningfully original and explicitly exclude mass-produced, repetitive material — which is exactly what an unsupervised pipeline produces, and the rejection arrives after you have done the work. And search visibility for generated content is a moving target that the platforms are actively adjusting against.

Content still works. It works when there is a person with a view in it, which is also the part that cannot be automated — so the economics are less "produce a hundred videos cheaply" than "produce ten good ones you could not have produced before".

Affiliate marketing

Recommend tools and earn a commission on signups. It pairs naturally with content: build genuinely useful guides and comparisons, include the links, and the earnings continue as the content keeps being found. This site does exactly that, which is why there is a disclosure on the relevant pages.

Three things decide whether it works honestly and lasts:

• Disclose, every time and visibly. This is a legal requirement in most markets, not etiquette, and burying it where nobody looks does not satisfy it.
• Recommend against things. A page that always concludes "buy it" is not trusted twice, and the trust is the asset. Saying plainly when a free option is enough is what makes the recommendation worth anything.
• Expect the commission to distort your judgement and build against it — decide what you think of a tool before you check what it pays.

Digital products, courses and small software

• Templates, prompt packs, ebooks — quick to produce, which means everyone can produce them, which means the market is crowded and prices are low. Works when it is attached to an audience that already trusts you.
• Courses — teach something you genuinely know. The saturated corner is courses about making money with AI, sold by people whose income is the course; be aware of which side of that you are on.
• A small tool — builders like Lovable or v0 get you to a working prototype without a team, which is real and new. The gap between a prototype and something people pay for and rely on is where most of these stop: support, billing, security, the edge cases, and the maintenance nobody budgeted.

The common failure across all three is building first and looking for buyers afterwards. The order that works is the opposite: find people with the problem, sell it before it exists if you can, then build it for them.

What to charge, when you have no idea

Under-pricing is close to universal at the start, and it is worse than it looks: a low price attracts the clients who are hardest to serve, and raising it later with the same client is awkward enough that most people never do.

Three ways to find a number without guessing:

• Work backwards from the saving. If an automation returns a day a month to someone whose time has an obvious cost, the value is visible to both of you. Ask what the process costs them now — most buyers know, and the number is usually larger than you expected.
• Work backwards from your year. Decide what you need to earn, divide by the weeks you will actually work, then by the hours you can realistically bill — which is roughly half the hours you work, once selling, admin and unpaid revision are counted. That figure is the floor, not the target.
• Ask the market. Freelance platforms show what comparable work sells for. Read what the higher-priced people say about themselves, not just the price.

Two habits that matter more than the number. Quote a fixed price for a defined scope, with revisions bounded — open-ended revision is where a profitable job becomes a loss. And raise the price when you are busy, not when you are desperate; a full pipeline is the only leverage that ever arrives on its own.

How to recognise the scams

This category attracts more deliberate deception than almost any other, and the patterns are consistent enough to list.

• Income screenshots. Trivially fabricated, never verifiable, and the only evidence offered in most of these pitches. Their presence is a signal in itself.
• Selling the method rather than using it. If the strategy worked as described, teaching it to competitors would be an odd business decision. Ask what the person earns from the thing itself.
• Urgency and closing doors. "Twelve places left" on an infinitely reproducible digital product is a manufactured pressure, and often an unlawful practice.
• Vagueness about the work. Real paths are describable in a sentence. "A system that leverages AI to generate passive income" is describing nothing.
• Upsells appearing after payment. The cheap course whose actual content is an advertisement for the expensive one.
• No way to fail mentioned. Every genuine business account includes the things that went wrong; a pitch with no failure in it is not an account of anything.

A reliable filter: would this person's advice still make them money if you succeeded? Someone teaching automation who also builds automations has aligned incentives. Someone whose entire income is teaching people to teach people does not.

What reliably does not work

Patterns that recur and fail, listed so you can recognise them before spending months:

• Reselling access to a model with a thin interface. The vendor's own product is cheaper and better, and they ship features faster than you can.
• A generic "AI agency" with no industry and no specific process. Nobody has the problem "I need AI".
• Publishing volume with nothing in it. The strategy that monetisation policies were written to exclude.
• Buying an "automated income system". If it worked as described, it would not be for sale.
• Waiting until you are ready. The first paid job teaches more than six months of preparation, and it can be small.

The unglamorous part

Whatever path you pick, once money changes hands it is a business, with obligations that differ by country and are worth getting right early rather than retroactively.

• Register and invoice properly, and declare the income. The rules for this depend entirely on where you are; find out rather than assume.
• Use a written agreement, even a short one: what you are delivering, when, for how much, who owns the output, and what happens if it stops working.
• Be careful with client data. If you paste a client's customer records into a tool, you have made a decision on their behalf. Check what the tool retains, and say so in your agreement.
• Do not promise ownership you cannot transfer. Generated logos, music and copy have unsettled ownership in several jurisdictions — relevant the moment a client asks for exclusivity.
• Keep your own records of what you used to produce client work, which model and when.

How to start — practically

1. Pick one path matching what you already know. Not the most lucrative-sounding one — the one where you have an unfair advantage.
2. Find one person with the problem before building anything. A conversation beats a business plan.
3. Do the first job cheaply and properly. Under-price the first one deliberately in exchange for learning and a reference.
4. Master the few tools that path needs — see the best AI tools — and ignore the rest.
5. Raise your price on the third client, and again after that. Nearly everyone under-charges for longer than they should.
6. Be consistent for longer than feels reasonable. Most of the people who failed at these paths stopped at the point where it was about to start working.

Transparency: nothing here is a promise of profit. Results depend on effort, skill, market and luck. Treat each path as a real business, because that is what it is.