Deepfakes on TikTok: What They Are, and How to Identify Them

There are all kinds of technologies that allow TikTok creators to go above and beyond when making their content. One of these is deepfake technology, which is where someone’s face (and sometimes their body as well) is digitally altered to appear as an entirely different person. Deepfakes are usually used for fun, but sometimes they’re used for harmful purposes such as identity theft or disinformation. As entertaining as they are, there have been increasing calls to crack down on anyone who uses them maliciously. 


How deepfakes are used for fun on TikTok

Most people love seeing clips of celebrities doing something hilarious. From trips and falls to wardrobe malfunctions, they show us that celebrities may be more similar to us than we thought. Thanks to artificial intelligence, though, humorous deepfakes of celebrities can be generated by the dozen. One account, @deeptomcruise, has amassed 3.6 million gained TikTok followers with deepfakes of Tom Cruise. Other notable names to be featured on deepfake accounts include Keanu Reeves, Robert Pattinson, and Mark Zuckerberg.

Why are these deepfakes enjoyed as entertainment, but others are frowned upon? It’s because that’s the whole point of these accounts – followers know that they’re looking at a fake video of a public figure, because the creator is open about it. Plus, none of the content reflects badly on the subject; it’s all made in good fun. 

In some cases, deepfakes are actually used for a more meaningful purpose. For example, some creators have “resurrected” historical reenactments and deceased actors, which gives us a closer look at the past. There are even people (usually not on TikTok) who use deepfakes as part of an online service, by recreating the faces of deceased loved ones to help with the grieving process.


How deepfakes are used for harm on TikTok

As you can tell, deepfake technology has a lot of great uses. Unfortunately, increasing numbers of deepfakes are being introduced on TikTok without people’s knowledge, for the purpose of harming either viewers, or the person who’s been digitally inserted into the video. The most common uses for this type of content are:

  • Spreading misinformation about electoral campaigns and political communication
  • Damaging the reputations of public figures
  • “Proving” fake news
  • Blackmailing or damaging the reputations of individuals by putting them in adult videos
  • Running identity scams and other frauds


Some of these are more dangerous than others. If a video of a celebrity saying something uncharacteristic emerges, it may get enough traction to be debunked. However, if scammers create a deepfake of a family member or friend asking someone for a wire transfer, it may work because the victim simply isn’t aware that the video could be fake. According to one study, identity frauds that involved digital manipulation cost $20 billion to US financial institutions in 2020.

While social media platforms such as TikTok, Meta, and more have taken steps to limit harmful deepfake content, it’s impossible for moderators to catch everything. A more effective approach is for users to identify deepfakes themselves, so they can stay savvy and report harmful content as needed. Whether you’re concerned about being scammed, or you just want to know what type of content you’re watching on TikTok, it wouldn’t hurt to figure out what the “tells” are in a deepfaked video. Here’s how you can do just that.


How to identify a deepfake

The first indication that you’re looking at a deepfake on TikTok is that something will just feel “off”. Even if you can’t put a finger on it at first, you may be able to tell that something’s up besides just one of the platform’s many filters. As you look more closely, you may notice the following things:

  • The hair doesn’t quite look realistic. Even though deepfake technology has become quite advanced, it still struggles with reproducing authentic hair textures and movements. This applies to both the hair on the head, and to facial hair.
  • The mouth and eyelid movements look unnatural somehow, like they’re forced or disjointed. These are parts of the face that often use very delicate movements, which deepfake technology can’t always replicate.
  • The face is too smooth. It may be ironic to point this out, given how many filters TikTok offers that can smooth out users’ faces, but this also applies to deepfake videos. Pay attention for a lack of wrinkling during facial expressions, or a lack of lines, pores, or general skin texture.
  • There are unusual shadows below the cheekbones. Even if you aren’t a film or lighting expert, you’ll probably be able to notice if the AI program failed to match the correct shadow angles to the deepfake.


In addition to examining the little details, you should also consider the broader context.

  • Does the TikTok account seem reliable and authoritative? Some accounts rely on sensationalist (and often false or exaggerated) videos to get more TikTok followers. If the account is full of tall tales and false claims, you probably shouldn’t trust any of their content.
  • Search elsewhere online to see if someone has already debunked the video. Many deepfake videos have already been debunked, and internet sleuths are constantly identifying new ones.
  • Evaluate the video with the subject’s character or reputation in mind. Do they seem to be contradicting themselves, or doing something else that’s out of character? Even videos that could be mildly controversial or damaging to their reputations should be suspect. Many public figures pay a lot of money to keep that kind of content off the internet, so if you can easily find one on TikTok, that’s cause for suspicion.


Deepfakes are a double-edged sword on TikTok, but it’s easier to spot them than you think

Artificial intelligence is advancing in all kinds of ways; some of them good, some of them bad. In the case of deepfakes on TikTok, they have some pretty great entertainment value, as well as a lot of potential for harm. As users become more aware of how to identify deepfakes, it’ll be easier to keep the good content and get rid of the bad.