Since the nascent days of the worldwide web, a popular joke has been, “If you saw it on the internet, it must be true.” Of course, this is sarcasm, as the web is full of inaccuracies, some intentional and some just outright mistakes. Furthermore, the advent of social media has greatly increased the reach of such falsehoods. But there’s a new player in the world of online shibai that’s making it harder, and this is known as ‘deepfake technology’ or ‘deepfakes.’

Deepfakes are videos created using artificial intelligence. As their name suggests, they are fake videos but highly believable. Such videos typically show celebrities, actors or politicians saying or doing things that never actually happened. The kicker is that not only is the video faked, so is the audio. AI is capable of re- creating actual human voices.

There are legitimate uses for deepfakes, such as special effects in movies. But the overriding concern is, as has always the case on the web, that they are used as part of a scam. Already, there are countless deepfakes depicting celebrities in pornographic videos. And while some might disagree, that’s not the worst of it.

Deepfakes are being used to spread disinformation and propaganda, both domestically and internationally. Celebrity endorsements of political candidates or political positions are being faked.

Government leaders are portrayed as saying things they’ve never said. Events that never happened are being presented as fact.

How can netizens determine whether what they are seeing is actually a deepfake? As of this writing, the giveaways are subtle, but the easiest way is to look at the subject’s face. While the main facial features are believable enough, the ancillary components often aren’t quite right. Hair (including facial hair), eyes, eyebrows, are often portrayed inaccurately. In many deepfakes the subject never blinks, or blinks incessantly. Lip movements don’t match the words, much like the old kung fu movies that showed in long-gone theaters of Chinatown.

In addition to facial giveaways, unnatural body movements are also indicators of a deepfake. It’s easier to make a deepfake if the body and head don’t move too much. Further, while voices can be faked, that’s a bit harder than video. As a result, the voice might seem a bit different from the actual subject’s voice.

When examining deepfake images, again, it’s the little details that give away the ruse. In images of a crowd, for example, look at the fringes. Oftentimes there are people with no faces, or two heads, or three arms.

Of course, the true method of determining whether you’ve come across a deepfake is to think critically. Don’t accept the first instance you see as gospel, especially if it is popularly controversial. Always look for confirmation from a markedly different source. Anything you see on social media should be immediately suspect.

John Agsalud is an IT expert with more than 25 years of information technology experience in Hawaii and around the world. He can be reached at jagsalud@live.com.