Modern phone scammers are increasingly using artificial intelligence: cloning voices, faking video calls and impersonating relatives, colleagues or officials. Just a few seconds of your video or a voice message may be enough to create a convincing “double”. However, even the latest AI systems have limitations, and you can use these to spot a scam in time.
Experts advise paying less attention to the quality of the voice or image and more to the caller’s behaviour and unusual glitches. Below are a few practical techniques that can help you detect a possible deepfake during a video call or phone conversation.
The “three-finger trick” helps you check a suspicious video call. If the person on screen introduces themselves as your relative or manager, ask them to hold up three fingers right next to their face or to change the angle by turning their head. For a real person this is simple, but many deepfake systems struggle to correctly process the face, fingers and background all at once in real time. Blurred fingers, flickering parts of the face or unnatural transitions can be a warning sign. At the same time, remember that this method is not a hundred per cent guarantee: similar “artefacts” can also be caused by a poor internet connection, so odd visual effects should be treated as a reason for further checks rather than as definitive proof.
During audio calls, the main task is not to determine whether the voice is “artificial”, but to break the scammer’s prepared script. Criminals usually act according to a pre-written plan, create a sense of urgency, try not to give you time to think and discourage additional questions. So you should do the opposite – ask unexpected questions and deliberately interrupt the conversation.
Specialists suggest asking about shared memories or details that have never been made publicly available. For example, you can ask about a family code word, childhood details or private family jokes. You should avoid information that is easy to find on social media – scammers often study Facebook, Instagram or TikTok profiles in detail.
Another tip is to deliberately pause and interrupt the caller’s monologue with everyday phrases such as: “Hang on, I’ll close the window” or “Wait a minute”. Someone close to you will take this calmly, whereas scammers often try to talk non-stop, ramp up the pressure and prevent you from collecting your thoughts.
A sudden change of subject is also useful. You can abruptly ask about a neighbour you both know, a shared celebration or a family event unrelated to the supposedly “urgent situation”. A real relative will respond naturally, while a fraudster will be forced to improvise or will try to quickly steer the conversation back to their script.
The simplest and one of the most effective checks is to hang up and call back yourself. In the conversation you can say directly: “I’m going to hang up now and call you back on your usual number”. A genuine relative or acquaintance will react reasonably. If the caller starts insisting that you do not end the call, invents new threats or ramps up the urgency, this is a strong reason to end the conversation.
In cases where someone claims to be calling from a bank, the police or another institution, experts advise asking for an official reference number and offering to call back via the general call centre or office number, which you will find yourself. Real employees will not object to such a check.
Specialists also warn: no matter how convincing a call may seem, certain information must never be shared over the phone. This includes one-time codes, online banking logins, PINs and passwords, passport and bank card numbers, as well as answers to security questions and confirmations of transactions in banking apps. Banks, government bodies and the police do not ask for this sort of information over the phone; if they do, you should end the call immediately.
Robust protection starts long before the first call. The less personal data are publicly available, the harder it is for scammers to construct a plausible story. It is useful for family members to agree on a code word that will serve as a kind of password known only to close relatives. This word will also come in handy if someone calls supposedly on behalf of a doctor, lawyer or law enforcement officer – a counter-question asking for the code will confirm or refute the authenticity of the call.
Experts also recommend reviewing your own “digital footprint”: open phone directories, public social media profiles and other services often contain exactly the sort of information that allows scammers to build convincing backstories. It is worth deleting unnecessary personal data and agreeing within the family on a simple rule: if in doubt, end the conversation and call back on a familiar or officially verified number.
Although artificial intelligence makes phone scams harder to detect, they are still based on the human factor: the scammers’ desire to control the conversation and to act under time pressure. You do not need to recognise the technology itself – it is enough to ask non-standard questions, interrupt the call or ring back yourself to spot the warning signs and protect yourself.