For years we told people to watch for bad grammar, odd phrasing, and generic greetings. Those tells are gone. Generative AI now writes fluent, personalised, convincing messages at scale — and it can clone a voice or a face from seconds of public video. Here is what changed, and what still works.
Voice cloning and the “grandparent” scam
A few seconds of audio — from a voicemail, a TikTok, a work webinar — is enough to clone someone’s voice. Scammers use it to call a parent or grandparent in a panic: “It’s me, I’ve been in an accident, please don’t tell anyone, I need money now.” The voice sounds real because, in a sense, it is.
Deepfake video and executive fraud
Businesses are being hit too. In one widely reported case, an employee wired millions after a video call with what looked like their CFO and colleagues — all deepfakes. “I saw them on video” is no longer proof.
Phishing without the typos
AI writes clean, on-brand emails in any language, and can tailor each one using details scraped from social media. The result: phishing that reads exactly like a real message from your bank, your boss, or a supplier.
What still works
- Verify on a channel you already trust. Hang up and call back on the number you have saved — never the one in the message.
- Agree a family code word. A simple shared word instantly exposes a cloned voice. Set one tonight.
- Slow down. Urgency is the one ingredient every scam still needs. Real emergencies survive a two-minute check.
- Assume voice and video can be faked. For anything involving money or credentials, confirm through a second, independent channel.
AI changed the surface of scams, not their engine. They still rely on pressure, secrecy, and a moment of trust. Slowing down and verifying independently still wins.
Not sure about a message right now? Paste it into ScamBite for a free, instant read on the warning signs.