How to detect DeepFakes on YouTube

In today’s digital age, where artificial intelligence (AI) is advancing at an incredible pace, distinguishing between real and fake videos has become increasingly difficult — especially on platforms like YouTube. Deepfakes — videos that are either altered or entirely generated by artificial intelligence to depict people saying or doing things they never actually did — have become a major cause for concern. These videos can be used for misinformation, political manipulation, financial scams, or even personal harm. I covered this in my previous article.

Understanding how to recognize deepfake content is not just a technical skill — it’s becoming a necessary part of being media literate in the 21st century.

• Pay Close Attention to Facial Details

One of the first places to look when evaluating a video is the person’s face. While AI-generated faces have become incredibly realistic, there are often small but noticeable inconsistencies. For example, the lighting on the face might not match the environment, or the skin tone may appear slightly off compared to the neck or hands. Shadows and reflections may also behave unnaturally — something that would be very rare in a genuine recording.

Another common issue is facial symmetry. Real human faces aren’t perfectly symmetrical, but AI-generated ones sometimes are, which can give them an uncanny, almost “plastic” appearance.

• Watch the Eyes and Eyebrows Carefully

The eyes are often referred to as the “windows to the soul,” and in deepfake videos, they can reveal quite a bit. Blinking patterns are a good indicator — real people blink regularly and naturally, while AI-generated characters may blink too much, too little, or not at all. The eyes themselves may lack depth or seem unusually still, without the subtle changes in pupil size or focus that occur in real life.

Eyebrows should move in sync with facial expressions. If someone smiles but their eyebrows stay completely still, or if their eyebrows appear pasted onto the face without blending naturally into the forehead, this could be a sign of a deepfake.

• Listen and Watch How Lips Move

Mismatched lip movement is one of the most common giveaways of a deepfake. In many cases, the mouth movements do not match the sounds being spoken — words may be pronounced before or after the lip movements, and sometimes the lips move out of sync with the words. Additionally, strange facial twitches around the mouth area are often noticeable. This is especially noticeable in videos where someone is speaking quickly or pronouncing certain sounds like “p,” “b,” or “m.”

Also, pay attention to the voice itself. AI voices can sound overly smooth, lack emotional inflection, or have unnatural pauses. If the speaker sounds robotic or seems to repeat phrases in a strange way, it could be a red flag. If a suspicious video features a public figure, you can review other videos of them, paying close attention to their vocal tone and speech patterns. 

• Be Aware of Accessories and Unique Features

Accessories such as glasses, hats, or jewelry can also help identify a deepfake. For instance, reflections in glasses may not match the surrounding environment or may be missing entirely. Hats or wigs might sit awkwardly on the head or fail to cast proper shadows.

Facial features like moles, scars, or tattoos are often altered or removed in deepfakes. If you’re watching a public figure and notice that a familiar mark is missing or appears in the wrong place, it’s worth questioning the authenticity of the video.

• Check the Source and Context

Just as important as visual clues is the context in which the video appears. Who posted it? Is it from a reputable news outlet or a lesser-known channel? Many AI-generated channels mimic documentary styles, especially in historical content. For example, some YouTube channels labeled as “historical documentaries” are fully generated by AI — complete with monotonous AI narration and oddly colored subtitles (often white or green).

These videos often claim to offer new insights into ancient civilizations, wars, or famous figures, but they’re created using synthetic images and voices.

Additionally, unverified sources often distribute videos while citing authoritative ones. In such cases, it’s crucial to verify whether the referenced organization actually published that content. For instance, if you encounter footage allegedly broadcast by a major TV network, check their official website and archives for such material.

When verifying context, do not overlook other sources of information, particularly checking for mentions or possible debunkings in the media. Additionally, if you encounter a video that raises suspicions, it is advisable to consult fact-checking resources, as they have likely already identified and analyzed it.

Conclusion

While it’s becoming increasingly difficult to recognize deepfake videos, cyber security providers are constantly improving their recognition algorithms. In this tug and war of who has a better tech game, it’s often citizens who are caught in the middle.

It is important to remember that despite the widespread use of AI-generated content, one should not engage in a “witch hunt” without thorough fact-checking beforehand. For example, there have been cases where a real video was dismissed as a deepfake, and many believed it simply because they assumed someone had already verified it. A striking illustration of this situation is the story covered in the GFCN article “Is it fake or a mistake? How the video with Trump in Saudi Arabia became a reason for speculation.”

However, you don’t need to be an expert to discern whether the video you are seeing is real.  

To begin with, be wary of the source. Trace the video back to where it was posted and by whom. You need to make a judgement on the legitimacy of the source, if it isn’t already explicit. Whether the source should be a fan account, or a news publication is often a good indication.  

Search engines are a powerful tool that can help you dig a little deeper and give you details like how far an image has traveled, unearth counterfeits, or show you where it originated from. Use search engines to do a reverse image check. In a world full of disinformation, staying informed and skeptical is no longer optional — it’s essential.