A video of a politician saying something outrageous. A voice message from your “daughter” begging for money. A photo of a celebrity in a place they never were. All fake — all increasingly convincing. Deepfake technology can now fabricate video, audio and images so realistic that experts struggle to tell the difference. Here’s what it is, how to spot it, and how to protect yourself and your family.
What Deepfake Technology Actually Is
“Deepfake” combines “deep learning” (the AI technique) with “fake”. The technology trains AI models on images, video or audio of a real person, then generates new content showing them doing or saying things they never did. Early deepfakes needed serious computing power and expertise; today’s tools run on phones, need only a few photos, and produce results in minutes. Voice cloning is the most accessible and dangerous variant — 30 seconds of someone’s voice from social media can generate convincing fake audio saying anything.
The technology itself is neutral: filmmakers use it for de-aging actors, educators for historical recreations, accessibility tools for voice restoration. But the malicious uses — fraud, political disinformation, non-consensual intimate imagery — dominate the headlines for good reason.
The Scams: What Criminals Actually Do
Forget Hollywood scenarios; the real threat is mundane fraud:
- The “family emergency” voice scam: AI clones a loved one’s voice — often from social media videos — then calls grandparents claiming arrest, accident or kidnapping, demanding immediate money transfer. This is happening now, in the UK, to real families. It’s the single most important deepfake threat for most people.
- CEO/business fraud: fake video or audio of executives instructing finance staff to transfer funds. Several multi-million-pound cases globally.
- Romance and investment scams: fake video calls building trust with victims before the financial ask.
- Non-consensual imagery: predominantly targeting women and girls — fake intimate images created from ordinary photos. UK law now criminalises creating and sharing these, but the harm is instant and the enforcement lags.
- Political disinformation: fake videos of politicians or events designed to mislead voters or inflame tensions — especially potent in the run-up to elections.
Our guide to AI and misinformation covers the broader information-warfare context.
How to Spot a Deepfake: Practical Tells
Detection gets harder every year, but current fakes still show signs:
Video tells: unnatural blinking (too much or too little), blurry or flickering edges around the face, mismatched lighting on face vs background, odd teeth or hands (AI still struggles with hands), jerky or inconsistent head movements, audio that doesn’t quite sync with lip movements, and a general “uncanny” quality that’s hard to define but real.
Audio tells: flat emotional affect, unnatural pauses or pacing, mispronounced names or local references the real person would know, background noise that doesn’t match the claimed environment, and urgency pressuring you not to verify (“don’t tell mum, just send the money now”).
Image tells: extra fingers or warped hands, garbled text in the background, inconsistent shadows, overly smooth “plastic” skin, and backgrounds that dissolve into nonsense on close inspection.
Context tells (often more reliable than technical ones): does the claimed event appear in reputable news sources? Does the video come from an official account or a random share? Is someone asking for money, personal info or urgent action? Would this person really communicate this way? Verification beats detection — you don’t need to analyse pixels if you can confirm through another channel.
How to Protect Yourself and Your Family
- Establish a family code word. Agree a secret word or phrase for genuine emergencies. If “your daughter” calls in crisis but doesn’t know the code word, it’s a scam. This single measure defeats most voice-clone fraud. Discuss it at the next family gathering — including with elderly relatives, who are prime targets (see our guide to tech for independent living for related scam awareness).
- Always verify through another channel. Got an alarming message? Hang up and call the person back on their known number. Got a shocking video? Check reputable news sources before sharing.
- Slow down. Every deepfake scam relies on urgency overriding judgment. A genuine emergency can survive a five-minute verification call.
- Limit public voice/video data. You can’t disappear from the internet, but consider privacy settings on videos featuring family members’ voices, especially children’s.
- Talk to teenagers specifically. Non-consensual deepfake imagery is a real and growing problem among young people. Our guide to talking to children about AI covers how to have this conversation.
- Report it. In the UK, report fraud to Action Fraud; report non-consensual intimate imagery to the police and the platform. The Revenge Porn Helpline supports victims.
What About Detection Technology?
AI detectors exist — tools claiming to identify deepfakes — but they’re unreliable, especially on new generation techniques, and they produce false positives that can discredit genuine content. Don’t rely on them for important judgments. The arms race favours generation over detection, which means human verification habits matter more than technology: verify identities, check sources, slow down.
The Bigger Picture
Deepfakes attack the foundation of shared reality: “seeing is believing” no longer holds. Societies will adapt — through media literacy, authentication technology (cameras that cryptographically sign photos), legal frameworks and new norms of verification — but the transition is rocky. The UK’s Online Safety Act and specific deepfake legislation are steps; enforcement and education must follow.
The Takeaway
Deepfake technology can now fake video, audio and images convincingly enough to power serious scams — especially voice-clone fraud targeting families. Protect yourself with a family code word, verify alarming messages through separate channels, learn the visual and audio tells, and never let urgency override verification. The technology will keep improving; your best defence isn’t detection software but the habit of pausing and checking. In the age of synthetic media, scepticism is a survival skill.
What Is Deepfake Technology? FAQs
How common are deepfake scams in the UK?
Increasingly common and underreported. Action Fraud data shows AI-assisted fraud rising sharply, with voice-cloning scams against elderly relatives among the fastest-growing categories. The technology that required expertise two years ago now runs on phones. Assume your family will be targeted eventually — preparation beats reaction.
Can deepfakes be detected reliably?
Not reliably enough to depend on. Detection tools exist but produce false positives and negatives, and generation technology improves faster than detection. For important judgments, verify through independent channels rather than analysing the media itself: call the person back, check official sources, slow down. Verification beats detection every time.
What should I do if someone makes a deepfake of me?
Document everything (screenshots, URLs, dates), report to the platform immediately, and report to the police — creating or sharing non-consensual intimate deepfakes is now a criminal offence in the UK. The Revenge Porn Helpline (0345 6000 459) supports victims specifically. Tell someone you trust; don’t suffer alone.
How do I protect my children from deepfake risks?
Two conversations: first, that fake images and videos of real people exist and are used to embarrass and bully — so they should tell you immediately if it happens to them or friends. Second, about their own digital footprint — photos and voice clips posted publicly can be source material. Privacy settings matter, but the deeper protection is a relationship where they’ll tell you when something’s wrong.
