Part 2 of 3: How to Catch an AI Scam in the Moment — Even When It Sounds and Looks Real

Part One of this series explained what AI scams are and how the technology behind them actually works. Now I want to get practical. Because knowing that AI voice cloning exists is not enough — you need to know what to do when you are in the middle of a call, and every instinct in your body is telling you it is real.

Here is what I want you to hold onto: the goal is not to become suspicious of everyone. The goal is to slow down long enough for your brain to ask a few quick questions.

Scammers need you to act before you think. The single most powerful thing you can do is pause. Any legitimate emergency can wait three minutes for you to make one phone call.

Step 1: Establish a Family Code Word — Before You Need It

This is the most important technique in this entire post, so I am leading with it.

Sit down with your family — your children, your grandchildren, anyone who might ever call you in an 'emergency' — and agree on a secret word or phrase. Something simple, memorable, and completely private. Maybe it is the name of a childhood pet. Maybe it is a family nickname or inside joke that never got posted online. It does not matter what it is, as long as only your real family members know it.

Then make this the rule: if anyone calls claiming to be a family member in a crisis, you ask for the code word before you do anything else. A real family member will know it immediately. An AI system operating from publicly available information will not — because it cannot access things that were never posted online.

This one step breaks the entire scam.

Step 2: Hang Up and Call Back on a Number You Already Have

This sounds almost too simple. But it works every time.

If you receive a distressing call from a family member — even if the voice sounds completely real — hang up and call them back directly on the number already stored in your phone. Not a number the caller gave you. Not a number a second 'lawyer' or 'officer' provides. The number you have always used.

If it was a real emergency, they will answer. If it was a scam, you just protected yourself from disaster.

The same rule applies to any call claiming to be from your bank, Medicare, Social Security, the IRS, or law enforcement. Hang up. Find the official phone number on their real website, or on the back of your card. Call that number yourself. Never trust a callback number provided by an incoming caller.

Step 3: Know the Four Fingerprints of Every AI Scam

AI scams — whether voice, video, or email — consistently share the same four characteristics. Once you recognize the pattern, the scam becomes much easier to identify no matter what form it takes:

• Urgency: You are pressured to act immediately, right now, with no time to think, verify, or call anyone else

• Secrecy: You are told not to tell family members, friends, your bank, or anyone who might talk you out of it

• Unusual payment: Gift cards, wire transfers, cryptocurrency, or cash delivered to a courier — always a method that cannot be reversed once it leaves your hands

• Emotional overload: The scenario is engineered to flood you with fear, love, or panic before logic can engage — a crying voice, a desperate story, a ticking clock

If a call or message hits even two of these four marks, stop. Hang up. Verify through a channel you control.

Step 4: On a Video Call, Look for the Seams

AI-generated video — sometimes called a deepfake — has gotten remarkably convincing. But it still has tells, especially in real-time calls where the AI has to work harder. Here is what to watch for:

• Blinking that looks slightly wrong — either too infrequent, too mechanical, or oddly timed

• Lip movements that are just slightly out of sync with the words, especially on harder consonant sounds

• A face that stays mostly front-facing and does not turn naturally to the side or look away

• Lighting that looks slightly different on the face than on the background behind them

• Skin that looks unusually smooth, or hair that seems a little artificial or blurry at the edges

• Pauses between sentences that feel slightly too deliberate, or transitions in speech that have a subtle mechanical quality

You do not need to be a technology expert to notice these things. You just need to trust your gut when something feels even slightly off. That feeling is information.

One practical test: ask the person on the video call to turn sideways and look at something off-camera, or to hold up both hands with fingers spread. AI video systems in 2025 and 2026 still struggle significantly with profile angles and with rendering hands accurately.

Step 5: Read Emails Completely Differently Now

The old advice about spotting scam emails — look for bad grammar, misspellings, awkward phrasing — is no longer reliable. AI has erased all of those tells. A scam email written with AI assistance reads exactly like a professional, legitimate message.

What to look for instead:

• The actual sending address: An email that appears to come from your bank but arrives from a gmail.com address, or from a name that is one letter off from the bank's real domain, is a scam — regardless of how polished the content looks

• Where links actually point: Hover your mouse over any link without clicking it and look at the web address that appears. If it does not match what the email claims, do not click it

• Any request for personal information: Your real bank, Medicare office, and Social Security Administration will never email you asking for your account number, Social Security number, or password — ever

• Unexpected attachments: Do not open any file attached to an email you were not expecting, even if it appears to come from someone you recognize

When in doubt, do not use the link in the email at all. Open a new browser tab, type the company's website address yourself, and handle whatever needs handling from there.

Step 6: Ask Something Only the Real Person Would Know

If you are on a call and you are still uncertain, ask the person something specific — something an AI operating from public information would have no way to answer.

Not: 'What is your mother's maiden name?' That might be findable online.

Instead, try something like: 'What did we eat at Christmas two years ago?' or 'What did you give me for my last birthday?' or 'What were we laughing about the last time I saw you in person?'

A real family member answers easily. An AI system will stumble, give a vague non-answer, or try to redirect the conversation back to the crisis at hand. That reaction tells you everything you need to know.

Your AI Scam Detection Checklist

Print this out and keep it near your phone.

• Pick a family code word this week and share it only with immediate family

• Rule: any 'family emergency' call gets the code word question before anything else

• Any family member call in crisis: hang up and call their real number yourself

• Any company or agency call: hang up and call the official number you look up independently

• Watch for the four fingerprints: urgency, secrecy, untraceable payment, and emotional overload

• On video calls: watch lip sync, look for unnatural blinking, ask them to turn sideways

• In emails: check the actual sending address and hover over links before clicking anything

• When uncertain: ask one specific question only the real person could answer

• Before sending any money for any emergency, call one more family member first, always

In Part Three, I cover what to do after a suspicious contact — and what immediate steps to take if money has already moved. Being scammed does not mean it is over. Some agencies can help, and can sometimes recover funds if actions are taken quickly enough.