The Receipts Index
How Scammers Use AI Now: Deepfakes, Voice Clones, and Fake Everything
The same AI tools that help people write, design, and translate also let scammers fake faces, voices, and writing cheaply and at scale. A convincing video of a famous person used to take a studio. Now it takes a laptop and a stolen clip of audio.
That breaks the old advice. People were told to spot the fake by bad grammar, weird lighting, or a robotic voice. Those tells are fading fast. What has not changed is the machinery underneath: false trust, fake urgency, then a request for money or access. AI upgrades the costume, not the con. Here are the four big AI-powered scams, the tell for each, and the defense that still works.
1. Deepfake videos of famous or trusted people
The pitch: a video of a famous billionaire, a celebrity, or a bank executive promoting an investment platform, a crypto giveaway, or a "limited" opportunity. The face is right, the voice is right, and it may appear in a hijacked or lookalike account. It always ends the same way: click this link, act before it closes.
The tell: a real opportunity is never a video telling you to act fast. Famous people do not double your crypto, and real institutions do not announce products through countdown timers.
The defense: never use the link in the video. A real offer would exist on the organization's official website, which you can reach by typing the address yourself. Check there. It won't be.
2. Voice clones of family members and bosses
The pitch: a phone call that sounds exactly like your child, your parent, or your boss. There's an emergency. An accident, an arrest, a wire that must go out within the hour. The voice may cry, panic, or pull rank, and it asks for money or gift cards while begging you to skip the usual process.
A few seconds of audio from a social media post is enough to clone a voice, and caller ID can be faked too. Sounding right is no longer proof of anything.
The tell: a voice you trust, plus an emergency, plus a request for money or secrecy, all in one call.
The defense: hang up and call the person back on the number you already have, not the one that called you. For family, agree on a code word in advance and treat any emergency call that can't produce it as fake. For work, verify every payment request through a second channel, no matter who appears to be asking.
3. AI-written phishing and romance messages
The pitch: emails, texts, and dating app conversations that used to be full of typos are now fluent, warm, and personalized. AI can read your public posts and mention your job or your city. A romance scammer can run hundreds of convincing conversations at once, in any language, for months, before the investment platform ever comes up.
The tell: there isn't a reliable one in the writing anymore. Perfect grammar proves nothing, and neither does a long, patient relationship.
The defense: judge the request, not the grammar. Whatever the message sounds like, ask what it actually wants: a click, a login, a payment. That request is the scam or it isn't. And keep one hard rule: never move money to someone you have not met in person, no matter how real they feel.
4. Fake customer support and fake AI assistants
The pitch: you post about a problem, and "support" finds you first. Or you search for a help number and land on a fake one. Or an official looking chatbot walks you through "verification." However it starts, it ends with a request for your password, your seed phrase, a one time code, or remote access to your device.
The tell: the ask itself. No real support team, human or AI, ever needs your password, your seed phrase, or the code that was just texted to you. Those exist so that only you have them.
The defense: only contact support through the official app or website you reached on your own, and refuse those three asks from anyone, every time. Polished AI chat changes nothing. The ask is the answer.
The one principle that AI cannot beat
AI can fake how something looks and how it sounds. It cannot fake a verified channel that you control. A scammer can clone a voice, but they cannot answer the real number you have saved. They can forge a bank's email, but they cannot change what you see when you type the bank's address yourself and log in.
So the durable rule is this: always verify through a path you chose, not a path they gave you. Every link, number, or chat window supplied by the other side is their territory. Leave it and re-enter through a door you already trust. Reported losses from AI-assisted fraud keep growing worldwide (the FBI's IC3 2025 Internet Crime Report, released in April 2026, flags AI-driven scams specifically), but you don't need the numbers to act. You need the habit.
If you want to build that habit with people who check receipts before money moves, the free Trunkline community is here: https://whop.com/alphavault-6be6/alpha-vault-founding-member
The AI-era screenshot checklist
Save this. Read it before anything urgent moves you.
- A video of a famous person urging you to act fast is a scam. Every time.
- Never use the link, number, or QR code the other side gave you. Find the official one yourself.
- Emergency call from a loved one or boss? Hang up and call back on the number you already have.
- Set a family code word now, before you need it.
- Perfect writing proves nothing. Judge the request, not the grammar.
- Never send money to or invest through someone you haven't met in person.
- No real support ever asks for your password, seed phrase, or one time code.
- When anything feels urgent, wait. Real matters survive a callback. Scams don't.
Real numbers. No hype. Receipts.