Spot AI phone scams right away with these tricks

In the past, phone scammers often gave themselves away with their poor English, a crackling line or implausible stories. Today, the situation is different: modern AI can mimic voices with uncanny realism and even manipulate live video calls.

Just a few seconds of audio from a video or a voice message is enough for fraudsters to impersonate relatives, friends or colleagues. As a result, shock calls, the classic ‘grandparent scam’, in which an apparent young relative in trouble begs for financial help, or supposed emergency calls seem more credible than ever.

The good news is that even modern AI has its limits. It not only has to imitate voices or faces, but also react spontaneously, calculate movement correctly and maintain a credible story. It’s precisely these weaknesses that you can exploit. Often, just a few seconds are enough to expose a scam attempt.

The 3-finger trick: How to expose many video deepfakes in seconds

If you receive a suspicious video call on your smartphone from someone claiming to be a relative, colleague or line manager, you can carry out a surprisingly simple test: ask the person on the other end to hold up three fingers right in front of their face.

For humans, this is a simple, everyday movement. However, many deepfake systems struggle with it. They have to process the face, fingers and background simultaneously in real time. This can result in blurred fingers, flickering parts of the face, unnatural transitions or brief glitches in the image. Even a quick turn of the head or an unusual camera angle can throw the fraudsters’ system off balance.

It’s important to note, however, that the ‘three-finger trick’ is not 100% accurate. Modern deepfake systems are improving, and even poor internet connections can cause image glitches. You should therefore treat visual oddities as a warning sign and look out for other clues. Plus, the test naturally only works for video calls – for audio-only calls, the following strategies will help.

Outwitting audio deepfakes: Throw fraudsters off their game

With audio-only calls, the same basic principle applies – but on a different level. You don’t need to identify that it’s an AI voice, but rather throw the caller off their prepared script. Whether it’s a cloned voice or a traditional telephone scammer makes little difference: criminals work with fixed scripts, they deliberately create a sense of urgency and rely on you asking as few questions as possible. It’s precisely this routine that you can disrupt.

Ask questions that can’t be Googled

Questions about a recent holiday or pet often aren’t secret enough these days. A lot of personal information can be found on Instagram, Facebook or TikTok. Memories or details that have never been made public are much more effective.

For example:

• What’s our family code word?
• Where’s the spare key?
• What was my first primary school teacher’s name?
• What nickname did grandad call you?

A cloned voice can be a perfect imitation. But the scammer has no knowledge of your personal memories.

Deliberately interrupt the flow of conversation

Fraudsters always want to maintain control of the conversation. Every pause gives you the chance to think, ask questions or assess the situation – which is exactly what criminals want to avoid.

So make a point of interrupting with an everyday remark such as, “Hold on, I need to close the window” or “Wait a moment”.

A real family member will wait patiently or react quite normally. Scammers, on the other hand, often try to carry on talking, to increase the pressure or to prevent you from interrupting.

Throw the conversation off course

Fraudsters rely on prepared stories and try to steer the conversation as closely as possible to their plan. You can disrupt this routine: suddenly change the subject or ask a question that has nothing to do with the alleged emergency.

Suitable examples include:

• ‘Has Grandma actually sold her car yet?’
• “How’s our neighbour getting on?”
• “What did we cook together last Christmas?”

Genuine family members will react spontaneously and naturally. Fraudsters have to improvise, evade the question or quickly try to return to their prepared story.

The most effective test: call them back yourself

An even simpler and often more effective approach is to ring them back yourself. The key phrase is: “I’m going to hang up now and ring you back on your usual number.”

This immediately robs fraudsters of their biggest advantage: control of the situation. A genuine family member will be understanding. Anyone who tries to stop you from hanging up by issuing new warnings, making excuses or putting you under additional time pressure is sending a clear warning signal.

Cheat sheet for spotting scams: These tricks will throw deepfakes off

This overview summarises the key tactics you can use to throw fraudsters off their prepared script and help you to better assess suspicious calls.

Tactic	Application	How it works
3-Finger Trick (Video)	Ask the person you’re speaking to to hold up three fingers directly in front of their face.	Many live deepfakes struggle to cleanly separate fingers and face in real time. Blurry fingers, flickering parts of the face or unnatural transitions can be warning signs.
Change of angle (video)	Ask the person to turn their head to one side or briefly show themselves from a different angle.	Some deepfake systems falter when faced with unusual perspectives or rapid movements, thereby giving themselves away.
Everyday question (audio)	Ask about a shared memory that has never been made public, such as: “What soap did you recommend to me when I was younger?”	The caller must react spontaneously and cannot rely on publicly available information.
Trick question (audio)	Ask about a deliberately made-up person or situation, such as: “How is Aunt Erna getting on, by the way?”	Genuine relatives would ask for further details. Fraudsters often unconsciously confirm made-up details to keep their story going.
Service number trick (public authorities)	When dealing with alleged police officers, bank staff or government officials, ask for an official reference number and say you’ll call back via the switchboard.	Genuine staff will have no problem with this. Scammers will try to prevent a call back by putting you under time pressure or making excuses.

Why these tricks work

All these methods target the same weakness: fraudsters need to control the conversation. They rely on time pressure and a prepared script. Every unexpected question, every interruption and every callback forces them to improvise – and this is often when crucial warning signs become apparent.

Never disclose this information over the phone

Even if the caller seems convincing or appears to know personal details, certain information should never, under any circumstances, be disclosed over the phone.

This includes, specifically:

• One-off codes
• Online banking login details
• PINs and passwords
• ID or credit card numbers
• Answers to security questions
• Authorisations in banking apps

Reputable banks, government agencies or police stations will never ask for such information over the phone. If you are asked for this, end the call immediately.

The best protection starts long before the first call

The less personal information that’s publicly available, the harder it is for fraudsters to come up with credible stories.

A shared family code word is particularly useful. It works like a traditional password and should be easy to remember but impossible for outsiders to guess. A random phrase such as ‘blue whale’ is better than the names of pets, birthdays or holiday destinations, which could potentially be guessed using social media.

The code word is also helpful if a caller pretends to be a doctor, solicitor or police officer. A simple counter-question such as, “What code word did my grandson give you?” puts a stop to many scam attempts right away.

It’s also worth taking a look at your own digital footprint. Public telephone directory listings, social media and publicly accessible profiles often provide fraudsters with exactly the information they need to craft convincing stories. You should therefore remove any unnecessary personal details and agree on a simple rule within the family: if in doubt, always hang up and call back using a number you recognise.

Conclusion

AI makes phone scams more dangerous, but not unbeatable. Whether it’s voice cloning or live deepfakes: ultimately, there’s still a human being behind the scam, working under time pressure and relying on the conversation to go as smoothly as possible.

That’s precisely why you don’t need to recognise AI – you just need to throw the scammer off their game. The three-finger trick during video calls, asking personal follow-up questions, deliberately interrupting the conversation, and calling back via a known number are often enough to throw even deceptively realistic AI fakes off balance.

Modern smartphones now offer additional protective features such as scam warnings, spam filters and security functions that can detect suspicious calls at an early stage. If you’re thinking of changing your device anyway, our smartphone buying guides – best phones, best mid-range phones and best budget phones – will help you choose.

• Yes, there’s another heatwave coming. But no, don’t put your phone in the fridge
• The Pixel 11 series is coming soon. Here’s what you need to know