AI Scams and Deepfake Protection: A Practical Guide to Spotting Fakes and Staying Safe

Artificial intelligence did not invent fraud. It industrialised it. The family-emergency call, the too-good-to-be-true store, the flawless email from “your bank” — all of these existed before. What generative AI changed is the cost of making them convincing: a loved one’s voice from a short clip, a famous face endorsing a product it never touched, perfect grammar in any language, a complete fake shop built in an afternoon. The US Federal Trade Commission reported that consumers lost more than $12.5 billion to fraud in 2024, a 25% increase on the year before — and the FBI’s Internet Crime Complaint Center (IC3) has formally warned that criminals now use generative AI to do it. This guide maps the four scams you are most likely to meet, and the protection system — habits, not software — that defeats all four.

Scam one: the voice-clone family emergency

The phone rings and a familiar voice is in trouble — a car accident, an arrest, a voice shaking with panic. The FTC’s consumer alert “Scammers use AI to enhance their family emergency schemes” explains the mechanic: a scammer needs only a short audio clip of a person’s voice — easily harvested from content posted online — and a voice-cloning program to sound exactly like your child or grandchild. The FBI’s PSA describes criminals generating short clips of a loved one’s voice to impersonate a close relative in a crisis, asking for immediate financial assistance or demanding a ransom. The ask always arrives with two accessories: urgency and secrecy — and a payment method that cannot be reversed, such as wire transfers, cryptocurrency, or gift cards and their PINs. Our dedicated guide covers the full defence: voice clone scams and the family code word that stops them.

Scam two: deepfake video calls and fake celebrity endorsements

Video used to be the end of an argument. No longer. The IC3 warns that criminals use AI-generated video in real-time chats with people posing as company executives, law enforcement or other authority figures, and use AI images of celebrities and social personas to promote counterfeit products and non-delivery schemes. The FBI’s practical advice is to watch for the seams — distorted hands and feet, unrealistic teeth, eyes and accessories, inaccurate shadows, lag, and voices that do not quite match the face — while accepting that the seams are closing. The durable test is procedural: end the call, find the organisation’s genuine contact details yourself, and verify there. We break down the checks in deepfake video calls and fake celebrity ads. Note what this scam is not: it is not a matter of spotting fakes for entertainment — it is a payment decision, and the payment is the point where you still hold all the power.

Scam three: phishing that reads perfectly

The typo-riddled phishing email is becoming a period piece. The FBI notes that generative AI corrects the human errors that might otherwise serve as warning signs of fraud, and that AI-written text now powers fictitious social profiles, content for fraudulent investment websites, and chatbots planted inside scam sites to talk victims through their doubts. When the writing is perfect, the tests have to change: verify through a channel you open yourself, judge the request rather than the prose — no legitimate organisation demands gift cards, wires or same-day secrecy — and heed the FBI’s subtler clue that tone and word choice can still betray a fraud. Our step-by-step version: AI phishing emails and texts, and how to check them. If you want to understand why even honest AI output needs this treatment, our guide on AI myths and scams covers the beliefs that leave people exposed.

Scam four: the fake shopping website

A polished store, professional photos, five-star reviews, a helpful chatbot — and nothing behind it. The IC3 says criminals generate content for fraudulent websites and images for non-delivery and counterfeit schemes. What a fake store cannot generate is a track record. The FTC’s online shopping guidance supplies the tests: search the store’s name with “complaint” or “scam”; read full descriptions, total costs and return terms; know that sellers must ship within the time promised — or within 30 days if none is stated; remember that https proves encryption, not honesty; pay by credit card for its dispute rights; and never buy from sellers who accept only gift cards, wire transfers, payment apps or cryptocurrency. The full routine is in our pre-purchase checklist for AI-era stores.

The protection system: five habits that cover all four scams

• 1. Agree a family code word. The FBI’s IC3 advises creating a secret word or phrase with your family to verify identity. Any “emergency” call that cannot produce it ends immediately.
• 2. Always call back on your own number. The FTC: don’t trust the voice — call the person on a number you know is theirs; if they are unreachable, go through another family member or their friends.
• 3. Verify every request out of band. The FBI: hang up, research the organisation’s real contact, call directly. Applies to calls, video, email and text alike.
• 4. Control the payment rail. Credit cards for purchases; never wires, gift cards, payment apps or crypto to anyone who contacted you first. The FTC notes these are the methods where recovery is hardest.
• 5. Shrink your raw material. The FBI advises limiting online content of your image and voice, keeping social accounts private and followers to people you know — less audio and video to clone.

If it happens anyway: the first 24 hours

Even careful people get caught on a bad day, and the FTC’s guide “What To Do if You Were Scammed” is built on speed: contact the bank, card issuer, app or wire company immediately and ask for a reversal; secure any account whose password was shared; and if your Social Security number was exposed, start a recovery plan at IdentityTheft.gov. Then report — at ReportFraud.ftc.gov, where the FTC says reports help it build cases and spot trends, and at IC3.gov, the FBI’s portal for internet crime, including the transaction details the FBI asks for. Our timed walkthrough is here: scammed by an AI fake — your first 24 hours.

What this means

Read the official guidance side by side — the FTC’s consumer alerts, its shopping and scam-recovery guides, the FBI’s generative-AI announcement — and a striking pattern emerges: almost none of it asks you to become a better lie detector. The agencies that investigate these crimes have concluded that detection is a losing game for ordinary people, because the fakes improve monthly and attention does not. Their entire strategy is to move the contest to ground where the scammer cannot follow: a word that exists only in your family’s memory, a phone number you saved years ago, a card network’s dispute process, an official portal. Fraud has always been a confidence trick; AI made the confidence cheaper to produce. Verification is how the price goes back up.

For US readers, the comparison with the pre-AI era is instructive. The old advice — check the spelling, look for the padlock, trust your ears — treated your senses as security instruments. They were never very good ones, and now they are obsolete ones. What replaces them is closer to how institutions already protect money: dual control, callbacks, separation between the request and the approval. Households can run the same system at kitchen-table scale, and the cost is one conversation. The stakes justify it: $12.5 billion in reported losses is certainly an undercount of the true figure, since the FTC’s own data book is built from the minority of victims who report. Every family that sets up the five habits does double duty — it protects itself, and it protects the relatives whose voices and faces are now raw material. AI has made believing your eyes and ears optional. Verifying is not.

One audience deserves a specific mention: children and older relatives, who sit at opposite ends of the same risk. Teenagers produce the most cloneable material — public videos, voice notes, livestreams — while grandparents hold the savings and the instinct to rescue first. The five habits connect those two facts. A teenager who keeps accounts private starves the clone; a grandparent who insists on the code word defeats it. Neither needs to understand the technology. They need the family to have had one unglamorous conversation — ideally this week, at the kitchen table, before anyone’s voice arrives on a phone asking for help.

Related Guides

• Voice Clone Scams: How They Work and the Family Code Word That Stops Them
• Deepfake Video Calls and Fake Celebrity Ads: How to Spot Them Before You Pay
• AI Phishing Emails and Texts: Why They Read Perfectly Now — and How to Check Them
• Fake Shopping Websites: A Pre-Purchase Checklist for AI-Era Stores
• Scammed by an AI Fake? Your First 24 Hours and Where to Report in the US

Bottom line

The agencies that fight these crimes have stopped asking you to spot fakes. Move the contest to ground the scammer cannot reach: a family code word, a call-back on a known number, a second channel for every urgent request.

Sources

• FBI Internet Crime Complaint Center — Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (PSA I-120324-PSA, 3 December 2024) (accessed 7 November 2026)
• US Federal Trade Commission — Scammers Use AI to Enhance Their Family Emergency Schemes (accessed 7 November 2026)
• US Federal Trade Commission — New FTC Data Show a Big Jump in Reported Losses to Fraud to $12.5 Billion in 2024 (accessed 7 November 2026)
• US Federal Trade Commission — Online Shopping (accessed 7 November 2026)