The short answer
No. Not in the way the adverts mean it, and not because the technology has yet to catch up. The interesting part is why. The difficulty is not that the models are too small, the data too thin or the computers too slow. It is that forecasting prices is a fundamentally different shape of problem from the ones AI is extraordinary at, and no amount of progress on the second kind of problem turns into progress on the first.
Before going further: this article is educational. It is not financial advice, it does not recommend or discourage any investment, and nothing in it is a suggestion to buy, sell or hold anything. If you are making real decisions with real money, speak to a qualified adviser who is regulated in your own country.
What AI is genuinely brilliant at
Modern AI is remarkable at one particular kind of task: finding patterns in a large pile of examples, where the rules producing those examples stay put.
Reading a scan works like that. Human anatomy does not rearrange itself because a model got good at spotting things. Translating between languages works like that. So does recognising a face, transcribing a recording, or predicting the next word in an ordinary English sentence. In each case the world holds reasonably still while the system learns it, and being right tomorrow means the same thing it meant yesterday.
That stillness is doing an enormous amount of quiet work in the background. Take it away and the whole approach changes character.
Markets do not hold still
Two things make prices different, and they stack.
The first is that prices move on information that does not exist yet. Whatever is currently known is already reflected in the price. What moves it next is the thing nobody has: an announcement, a resignation, a court ruling, a storm, a decision taken in a meeting this afternoon. No quantity of history contains tomorrow's news. A system trained entirely on the past is not being cautious about the future. It simply cannot see it.
The second is stranger, and it is the bit worth carrying away.
In a market, finding a pattern changes the pattern.
Imagine a rule that genuinely told you a price was about to rise. The moment anyone acts on it, they buy, which nudges the price up before the rise they were expecting. As more people act, the edge shrinks, until acting on it is no longer worth the cost of doing so. The pattern does not gently fade with age. It is eaten by the act of using it.
Nothing in medical imaging behaves like this. A tumour does not move because you found it. This is the single deepest reason to be sceptical of anything sold as a repeatable edge, and it holds whether the thing doing the finding is a person, a spreadsheet or the most capable model in the world.
The question that does most of the work
Which brings us to a question worth keeping in your pocket permanently.
If someone had a system that reliably turned money into more money, why would they sell it to you for a monthly fee?
Sit with that. Selling access is a peculiar business decision for anyone holding an actual money machine. Running it quietly would pay better, require no advertising, no customers, no refunds and no regulator. And if the edge is real, every subscriber makes it smaller for the seller. The business model tells you what the seller believes about the product.
There is a stock answer to this: they only want a handful of members, they are giving back, they want to democratise something. Notice that the reply is always about the seller's character and never about the mechanism. That is the tell.
What a backtest is, and why it is not evidence
Almost every impressive chart in these pitches is a backtest.
In plain English
- Backtest:
- A strategy run over old price data to show what it would have done. A rehearsal, not a result.
- Curve fitting:
- Tuning a rule until it fits past data beautifully, at the cost of it meaning anything about the future.
- Live track record:
- What happened to real client money, over a real period, including the bad months.
- Survivorship bias:
- Seeing only the winners, because the failures quietly stopped being mentioned.
The trouble is not only dishonesty, though there is plenty of that about. The deeper trouble is that anyone who tries enough variations on historical data will find some that look magnificent by coincidence alone. That is curve fitting, and it catches out people acting in complete good faith. A backtest is a rehearsal in an empty theatre where you already know every line, so it says very little about opening night.
Watch the tense in these adverts. "This strategy would have returned" is a completely different sentence from "our clients received", and the two get blurred deliberately.
How to read an AI money pitch
Once you have seen a few of these side by side, the striking thing is how unoriginal they are. The same moves recur across unrelated countries and brands, because they are tested templates. You do not need to understand markets to spot a template.
| What you are shown | What it actually is |
|---|---|
| A wall of winning screenshots | A selected sample, with the losing days not posted |
| "Would have returned" figures | A backtest built with hindsight, over data already known |
| A countdown clock or a closing door | Pressure designed to stop you going away to think |
| Cars, watches, a view | Evidence of spending, which says nothing about returns |
| A famous face endorsing it | Often synthetic, and frequently denied by the person |
| "Guaranteed" or "risk free" | Language regulators in many countries watch closely |
| A move into a private chat | Fewer witnesses, and no record kept by any platform |
| A fee required to withdraw | Money going out, dressed up as money coming back |
That last one deserves its own warning.
Being asked to pay anything in order to release money that is supposedly already yours is the clearest single signal there is. Legitimate firms deduct charges from a balance. They do not ask you to send fresh money in.
You do not have to argue with anyone. You only have to ask a few dull questions and watch what happens to the conversation. Who is the regulated entity behind this, and what is its registration number? Is this real client money or a simulation? What were the worst three months? Who holds my money, and what exactly do I own? How would I withdraw all of it today?
Then look that registration number up yourself, on your own country's financial regulator's public register, by typing the regulator's address in rather than following a link you were sent. Rules and registers differ by country, so check the one that covers you rather than assuming something you read online applies.
The answers matter less than the reaction. Genuine firms answer dull questions dully. Pitches respond with flattery, irritation, a change of subject or a brand new deadline.
What AI actually does in real finance
It does a great deal. It is just much less cinematic than the adverts.
Banks and insurers use it to flag suspicious transactions, to read and sort enormous quantities of documents, to run compliance checks, to handle routine customer queries, to summarise long filings and reports, and to model risk. That last one is worth noticing, because risk modelling produces ranges and probabilities rather than forecasts. It is a tool for describing how wrong things could go, not for telling anyone what happens next.
Notice what unites all of it: processing, sorting, checking and summarising. That is pattern work on data where the rules hold still, which is exactly what these systems are good at.
What it can do for your own money
Genuinely useful, genuinely boring, and available to anyone.
It is good at explaining a confusing statement, letter or contract in plain English. It is good at translating jargon you were too embarrassed to ask about. It can help you draft a complaint to a company and keep it polite and factual. It can organise and categorise spending data you have already exported yourself. And it is very good at preparing you for a meeting with a real, regulated adviser, by turning a vague worry into a list of specific questions.
These tools are unreliable at arithmetic. They produce numbers that look plausible rather than numbers that are correct, so check every figure yourself. Use them to understand and organise, and use a calculator or a spreadsheet to count.
Where to go next
If this was useful, the full version is free. Our AI and Money course goes properly into why prediction is hard, how the pitches are built, what firms actually do, and the safe everyday uses. Since almost every technique in these pitches is a general scam technique wearing a financial costume, AI Safety, Privacy and Verification is the natural companion, particularly if a convincing celebrity video is the thing that made something feel real to you.
And if you want more on the gap between what AI promises and what it delivers, What Is the AI Bubble? covers the same instinct applied to the industry as a whole.