What is Deepfake and How to Recognise It?
The development of artificial intelligence has made the creation of photos, audio, and video increasingly easy. On one hand, this technology can be used for entertainment and creative needs. On the other hand, the same technology can be utilised to create fake content that appears to depict real events.
One term that has frequently emerged in such circumstances is ‘deepfake’. This content can make a person appear as though they are saying or doing something that never actually occurred. So, what is a deepfake and how can it be recognised?
What is a deepfake?
Deepfake refers to content in the form of images, videos, or audio that is created or manipulated using artificial intelligence and machine learning technology. This technology can be used to alter a person’s face, mimic voices, or make someone appear to perform actions they never actually undertook.
Its increasing sophistication means such content is not always easy to recognise at a glance. In fact, several studies show that humans still find it quite difficult to distinguish AI-generated content from authentic content.
Deepfakes are also beginning to be utilised in various modes of fraud. The FBI’s 2025 Internet Crime Report noted 22,364 complaints involving AI, with losses amounting to nearly $893 million USD. Perpetrators use fake profiles, voice clones, identity documents, and even videos featuring public figures or close acquaintances in a convincing manner.
How to recognise deepfakes?
- Pay attention to lip movements and audio
One thing that can be checked is the synchronisation between the voice and the movement of the mouth. In manipulated content, lip movements sometimes do not truly follow the audible sound.
Also, look for audio that sounds too flat, strange changes in intonation, or specific pronunciations that feel unnatural. However, such signs do not always appear as AI technology continues to evolve.
The FBI’s 2025 report also noted that in several cases of voice deepfake usage, the movements of the person appearing on camera did not fully align with the audio, including instances involving the sound of coughing or sneezing.
- Examine the face
The face can be one of the parts to inspect when viewing videos or photos suspected of being manipulated. Some signs to look for include unnatural blinking, overly smooth facial features, strange skin tones, or an unnatural boundary between the hair and the background.
Nevertheless, visual signs alone are not sufficient to confirm whether content is a deepfake. Camera quality, lighting, video compression, and recording conditions can also make a face appear unusual.
- Check the context and the source
Do not merely look at the content of the photo or video. Also, observe who first uploaded it, when the content was created, and whether there are other reliable sources discussing the same event.
If a video shows a famous figure making a shocking statement, do not immediately assume it is true just because the face and voice look convincing. Seek statements from official accounts or trusted media to verify if the information is correct.
This method is important because AI technology can create visually convincing content, while the surrounding context may indicate discrepancies.
- Beware of scams
Deepfakes are not only used to create entertainment content or spread false information. This technology can also be used as part of fraudulent schemes.
For example, someone might receive a video call or a voice message that appears to come from a family member, a superior, a public figure, or a specific party, and is subsequently asked to urgently send money or personal information.
In such situations, do not immediately comply with the request. Contact that person through a known and verified communication channel to confirm their identity.
The FBI also warns the public not to rush into fulfilling suspicious requests and to take time to verify information before handing over money or personal data.
- Do not rely on eyes and ears alone
Recognising deepfakes is becoming increasingly difficult as technology quality continues to advance. Therefore, a single sign is not enough to conclude that content is fake.
Cybersecurity agencies are also continuously developing technology to detect manipulated media. The National Institute of Standards and Technology (NIST), for instance, is developing a forensic evaluation programme to test detection technologies for deepfakes and AI-generated media.
For internet users, the safest step is to combine visual inspection, source verification, cross-referencing information, and maintaining a cautious attitude towards shocking content.
Deepfakes are not inherently bad. Digital manipulation technology can be used for creative, educational, and entertainment purposes. The problem arises when this technology is used to deceive, spread misinformation, defame, or harm others.
Therefore, when you encounter a photo, video, or audio that feels both too convincing and suspicious, do not share it immediately. Take the time to check the source and context first. Amidst the rapid development of AI, the ability to verify information is becoming increasingly vital.