AI scams: how to spot them and protect yourself

An AI scam is a classic con (an emergency, a fake relative, a fake adviser, a fake investment) made more convincing by artificial intelligence: a cloned voice, a doctored video, a flawless message. A few habits are enough to protect yourself: slow down, check through another channel, and never pay or share a code under pressure.

This guide helps you spot risky situations and respond. It does not explain how these frauds are built, only what you need to know to stay safe.

What AI changes about scams

The levers of fraud haven’t changed: fear, urgency, trust and the lure of easy money. Generative AI, which produces text, images, voices and video, doesn’t invent new emotions to exploit. It simply makes the trap more convincing, quicker to produce and easier to personalize.

In practice, three old warning signs no longer work:

• Spelling mistakes. A chatbot writes flawless messages in any language.
• A familiar voice. Text-to-speech tools can imitate a voice from recordings posted online.
• Video as proof. A deepfake can show a real person saying things they never said.

These scams are summed up in our glossary entry on AI-powered scams.

The six most common situations

1. The call from a loved one in trouble

A call or voice message sounds exactly like your child, partner or parent. They mention an accident, an arrest or trouble abroad, need money right now and ask you to keep quiet. We cover this scenario in detail in our guide to protecting your family from AI voice cloning.

2. The fake boss or fake supplier

At work, an email, a call or even a video meeting seems to come from a senior executive or a regular partner. It asks for an urgent, confidential transfer, or announces new bank details. Businesses of every size are targeted.

3. The celebrity investment tip

A video on social media shows a famous person praising a « guaranteed » investment or a miracle cryptocurrency. The video is fake, the platform is fraudulent, and the first « gains » you see are there to make you invest more.

4. The perfect message from your bank or a government agency

A flawless email or text, with your name and sometimes your employer or a recent purchase, asks you to click to « secure your account » or « avoid a fine ». It’s phishing, just better written than before.

5. The fake AI assistant

An app, browser extension or website imitates a well-known assistant, or offers a « free premium » version. Its goal is to steal your login details or card number, or to install malware. Only download assistants from the provider’s official website or the official app stores.

6. The online relationship that seems too good

An attractive profile, polished photos, weeks of warm conversation, then a request for money to cover an emergency. The photos and some of the messages may be AI-generated. If you have never met the person and money comes up, treat it as a major warning sign.

Check yourself A flawless text from your bank asks you to confirm a transaction by sharing the code you are about to receive. What do you do?

You never share that code with anyone. Your bank will not ask for it. Ignore the link in the text, call your bank on the usual number (the one on your card or in the official app), and report the message.

Warning signs to remember

Since the quality of the message no longer gives the scammer away, look at the situation instead. The table below matches each warning sign with the right response.

Warning sign	What it often hides	The right reflex
Extreme urgency, « right now »	Stopping you from thinking	Slow down, take ten minutes
A request for secrecy	Cutting you off from family or colleagues	Talk to someone you trust
Unusual payment (gift cards, crypto, new bank details)	Money that is hard to recover	Refuse, check through another channel
A request for a one-time code	Taking over your account	Never share it
Unexpected contact asking you to act	Impersonation	Call back on a number you know
Fast, guaranteed returns	Fake investment	Check the platform with the regulator

Should you rely on detection tools?

Some tools claim to tell whether an image, text or voice was made by AI. These AI content detectors can offer a clue, but they get it wrong both ways: they miss fakes and sometimes flag genuine content. Some providers also add an invisible watermark to the images or audio they generate, but scammers don’t necessarily use those tools, and a watermark can be lost after editing.

So don’t count on a tool to decide in the heat of the moment. A three-minute call to the right person beats any analysis. Keep one simple question in mind: « If this message were fake, what could I lose? » If the answer is money, codes or personal data, check before you act.

Everyday habits that work

With your family

• Agree on a family code word: a word or phrase only your close family knows, to ask for if a call worries you.
• Always call back on the number you already have, never the one the caller gives you.
• Talk about it with older and younger relatives: older people are often targeted by fake-relative calls, younger ones by fake contests and offers. Our guide to kids and AI has ideas for starting the conversation.
• Never blame a victim. These scams are designed to fool careful, well-informed people. Someone who feels ashamed may stay silent and lose precious time to block a payment. Tell your loved ones clearly that they can call you at any hour, even over a simple doubt.

With your accounts

• Turn on two-factor authentication for your email, social media and bank.
• Use a different password for every site.
• Never give out a verification code, even to someone claiming to be your bank, the police or tech support.

At work

• Use dual approval: no transfer or change of bank details based on a call, email or video meeting alone, even from the CEO.
• Verify through an independent channel: call back using the internal directory, or get a colleague to confirm.

With what you post

Scammers use what they find online: videos with your voice, family photos, details about your job and travel. Limiting what you make public reduces their raw material. That idea runs through our guide on AI and privacy: what happens to your conversations.

AI agents can be tricked too

If you use an assistant that reads your email or an agent that browses for you, it can be manipulated itself. A booby-trapped message or page may hide instructions it could follow: this is called prompt injection. Stay in control of sensitive actions (payments, sending, deleting) by requiring confirmation. Before delegating a task, our Can I hand this to an agent? tool helps you gauge the risk.

Likewise, don’t take a chatbot’s word for a customer service number or a website address: it can be wrong. Our guide on how to fact-check AI answers shows how to trace information back to the official source.

If you are a victim or have a doubt

Don’t stay alone with it and act fast. Here are the steps, based on official resources checked on October 22, 2026.

1. Contact your bank right away if you paid or shared card details, to try to stop or dispute the transaction.
2. Keep the evidence: phone numbers, messages, emails, screenshots, transfer references.
3. In the United States, report it to the FTC at ReportFraud.ftc.gov, the Federal Trade Commission’s site for reporting scams and fraud.
4. In the UK, the National Cyber Security Centre explains how to forward suspicious emails to its reporting service. It directs fraud victims in England, Wales and Northern Ireland to Report Fraud (reportfraud.police.uk or 0300 123 2040); in Scotland, victims should call Police Scotland on 101.
5. Change your passwords if you typed your login details into a fake site, starting with your email account.

This is general information. For a specific situation, such as a dispute with your bank, speak to a legal professional or a victim support organization.