← All guides Safety

AI images in crypto and investment scams

Fake trading screenshots, invented offices and generated founders. Which parts of an investment pitch a pixel check reaches, and which need a register lookup instead.

· 10 min read · Best AI Image Detector

Check the screenshots, not the founders. Fabricated profit screens and account balances are the images that actually persuade people, and they are the easiest of the set to test.

The four images in a typical pitch

Investment fraud is a genre with conventions, and its imagery is remarkably consistent. Four kinds of picture do almost all of the work, and each fails a different check.

  1. The profit screenshot. An account balance, a trading app, a chart in green. This is what converts curiosity into a transfer.
  2. The lifestyle photograph. Cars, watches, a view from somewhere expensive. Usually stolen from an unrelated real account.
  3. The team page. Founders, advisers, an office. Increasingly generated, and the easiest of the four to test.
  4. The document. A certificate, a licence, a registration number on headed paper. Usually a real template with the details changed.

Notice that only two of the four are detection problems at all. The lifestyle photographs are genuine pictures belonging to somebody else, which is a reverse image search question, and the licence claim is a register question.

People check the founders because faces feel checkable. The founders are the least important image in the set, and a fabricated scheme can operate perfectly well with entirely real, entirely unwitting photographs on its team page.

Pixel check Reverse search Register lookup
Profit screenshot Partly Screenshots degrade badly No No
Lifestyle photograph No A real photo, stolen Yes Finds the owner No
Team and office images Yes Its best case Partly Sometimes Partly Directors are public
Licence or certificate Partly Altered fields show No Yes This is the answer
Which check reaches which image. The one that persuades people is the hardest to test.

Why the profit screenshot resists checking

A screenshot is a fresh image made by your device from what was on the screen. That single step replaces the original encoding, removes sensor noise entirely and re-renders text through the display pipeline.

Everything a pixel model reads is downstream of that. A screenshot of a genuine trading app and a screenshot of a fabricated one both look, statistically, like screenshots, which is why scores on this category cluster in the middle and stay there.

The useful checks on a profit screen are therefore not about pixels at all. Do the numbers arithmetic correctly? Does the interface match the current version of the app it claims to be? Are the fonts and spacing right? Does the timestamp agree with a day the market was open?

Fabricated screens fail on arithmetic more often than on rendering. A balance that does not equal the sum of its positions, a percentage that does not match the figures beside it, or a date that falls on a weekend are all more reliable than any score.

Where the founders check does help

Team pages are the one part of the genre where detection is genuinely strong. A page of generated portraits produces high scores with flat maps across several images at once, which is a distinctive pattern.

Run three or four of the photographs rather than one. Real team photography, even bad team photography, varies: different days, different light, someone photographed against a window. Generated sets are uncannily consistent.

Then look the people up. Company directors are a matter of public record in most countries, professional registrations are searchable, and a founder with a decade of claimed experience who exists nowhere except one website is the finding.

  1. Look up the entity, not the images

    A company register, a financial regulator warning list, and the regulator own register of authorised firms. This answers the question the pictures were designed to distract from.

  2. Reverse image search the lifestyle photos

    A hit under another name in another country is common and settles the character of the operation immediately.

  3. Run a check on the team page

    Several portraits at once. Look for uniformly high scores with flat maps across the set.

  4. Audit the screenshot arithmetic

    Do the numbers add up, does the interface match the real app, does the date fall on a trading day.

  5. Test whether money comes back

    A small withdrawal request is the single most informative action available, and platforms running this pattern block it.

The pattern that no image can fix

Every scheme in this category eventually meets the same test, and it has nothing to do with pictures. Money goes in easily and does not come out. Deposits are instant; withdrawals require a fee, a tax payment, an upgrade or a verification step that never completes.

Anybody weighing an opportunity can run that test cheaply. Deposit a small amount, then immediately attempt to withdraw it. A genuine platform returns it; a fraudulent one produces a reason why not, and that reason is the entire answer.

This matters because it removes the burden of being right about images. You do not have to out-analyse a fabricated screenshot if you can simply ask for your money back and observe what happens.

Why generated imagery made this cheaper to run

The economics are worth understanding, because they explain the volume. Assembling a convincing investment operation used to require stolen photographs, which carried a risk: the real owner might find them, and a reverse image search might too.

Generated founders remove that exposure entirely. A team of six people who do not exist cannot complain, cannot be found in an earlier photograph, and can be regenerated the moment a site is taken down and rebuilt under another name.

The same applies to offices, certificates and event photographs. What used to be the expensive part of a fabricated business is now the cheapest, which is why these operations now appear with a full complement of imagery from the first day.

What has not become cheaper is the regulatory record. Authorisation, a company registration with a history, and directors who exist in public filings still cannot be generated, which is why those remain the checks worth running first.

Questions people ask

Can you detect a fake trading screenshot?
Poorly, and it is worth being direct about that. A screenshot re-renders everything through your display and removes the sensor noise a model reads, so genuine and fabricated screenshots both score like screenshots. Check the arithmetic, the interface version and the date instead.
What about the founders on the team page?
That is where a pixel check performs best. Run three or four portraits together and look for uniformly high scores with flat region maps across the whole set, which is the signature of generated team photography. Then look the individuals up in a company register.
The lifestyle photos looked real. Do they mean anything?
They are usually genuine photographs, taken from somebody else entirely. A reverse image search often finds the real owner, frequently under a different name in another country. A clean detection result on these is expected and means nothing.
What is the single fastest check?
Search your financial regulator register of authorised firms, and its warning list of unauthorised ones. It takes under a minute, it reaches the actual question, and it does not depend on being right about any image.
They sent a licence certificate. Is that verifiable?
Verify the number against the regulator register rather than examining the document. Certificates are usually real templates with details changed, which shows as one hot tile on a region map, but the register lookup is faster and conclusive.
How do I test a platform without risking much?
Deposit a small amount and immediately request a full withdrawal. Genuine platforms return it. Fraudulent ones produce a fee, a tax or a verification step that never completes, and that response tells you everything the images were designed to obscure.