Safe Retire Watch logo

How can seniors identify AI scams?

Short answer

Seniors should identify AI scams by structure, not by sound. Cloned voices and AI-written emails are now too convincing to detect by ear or by spelling. Instead, treat any contact combining urgency, secrecy, and an irreversible payment method as fraudulent, and verify by hanging up and calling back on a known number or asking for a prearranged family safe word.

Reviewed and updated by the Safe Retire Watch Research Team.

Part of our AI voice & deepfake scams hub.

Why the old advice stopped working

For years the guidance was to look for broken English, odd grammar, and voices that sounded 'off.' Generative AI removed all three tells. Scam emails are now fluent and correctly formatted, fake websites are pixel-accurate, and a relative's voice can be reproduced from a few seconds of public audio. Detection based on polish is obsolete.

What has not changed is the underlying mechanics. Fraud still requires you to act quickly, act alone, and pay in a way that cannot be reversed. AI improves the costume, not the plot.

What are the signs of an AI scam?

  • A family emergency call where the voice is right but the number is unfamiliar and the caller keeps the conversation short before handing you to an 'official.'
  • Any instruction not to tell other family members, your bank, or your advisor.
  • Payment by gift card, wire, cryptocurrency ATM, or a courier collecting cash — the four irreversible channels.
  • A video call that is unusually brief, oddly lit, or where the person avoids turning their head or holding up an object on request.
  • A perfect message from a company you use that still asks you to click, call, or confirm credentials.
  • An investment or romance contact whose photos, story, and 'proof' all check out but who cannot appear live and unscripted.

How do I verify a suspicious message or call?

  1. 1End the interaction first. You lose nothing by hanging up, and pressure not to hang up is itself a red flag.
  2. 2Reach the person or company through a channel you already have: a saved contact, the number on your card, or the company's app.
  3. 3For messages, do not click; open the account separately and look for the same alert there. Real alerts exist inside the account.
  4. 4Run the text of a suspicious email through our free AI scam email explainer, or check a link with our link checker, before acting on it.
  5. 5Check unfamiliar numbers with the phone reputation checker rather than calling back blindly.
  6. 6When still unsure, call one trusted family member. Two minutes of a second opinion prevents most losses.

How do I reduce the material AI can use against me?

  • Set social media profiles to friends-only and limit public video that captures your voice.
  • Use the default voicemail greeting rather than a recording of your own voice.
  • Avoid posting travel plans, hospital news, or family milestones publicly — those become scam scripts.
  • Ask family not to tag you in public videos, especially ones with long stretches of speech.
  • Keep your phone number off public directories and marketing lists where practical.

Frequently asked questions

Can AI really copy my grandchild's voice?

Yes. Modern voice-cloning tools need only a few seconds of clear audio, which can come from a social video, a voicemail greeting, or a recorded 'wrong number' call. The FTC and FBI have both warned about AI-assisted family emergency scams. Because the audio is convincing, verification must come from a callback or a prearranged safe word rather than from listening.

Are AI-written scam emails harder to spot?

Considerably. The traditional clues — misspellings, awkward phrasing, mismatched fonts — are largely gone. Focus instead on what the message wants: a click, a call to a number in the message, credentials, or a payment. Verify by opening the company's app or typing its address yourself, since the request path is the part AI cannot legitimize.

Is there technology that detects AI scam calls?

Carrier spam filtering and reputation tools help by flagging suspicious numbers before you answer, and they are worth enabling. But no consumer tool reliably detects a cloned voice in real time. The dependable defense remains procedural: never act on an inbound call, always call back on a number you already have, and use a family safe word.

Sources and further reading

Guidance on this page is based on current advisories from these authorities.

More questions answered