As unfortunate as it is, the newest generation of AI videos is so good that it’s borderline impossible to tell when something has been faked. Instead, what users need to do now is consider the logic of what they’re seeing, consider its source, and finally, assume fake til proven real.
Consider the Logic
The AI boom has lead to a bunch of “ragebait” videos, videos designed to evoke a very strong emotional response, hitting the web. The extent of the emotions isn’t limited to just rage: like clickbait before it, AI ragebait videos are often offering a feeling of self-righteousness, disgust, or confusion. Some of them are even funny, at an AI avatar’s expense. It’s a callback to the engagement-farming (content meant to get clicks, comments, or downloads, via intentionally doing something annoying to watch in a video or ad) of earlier years. Since that earlier material was often obviously scripted, it was easy enough for a decently-sized group of people online to simply ignore it. Well, now the videos are good enough that those people who could previously rely on intuition no longer can.
For a little while now, TikTokers have recorded videos of themselves messing with people in public by putting a picture of them or their property into an AI app and telling it “generate a video of this truck being stolen” before showing it to a person who’s just trying to buy groceries and then go home as a ‘prank’. Obviously the guy goes out to check, interrupting his day, leaving his cart behind, all while someone is filming him, and when he sees the car is still there, the ‘joke’ seems to be that he got fooled at all by a video played on someone else’s phone. Notably, nobody is laughing in the video, including the person filming. The prankees are suddenly incredibly worried about their property or their reputation; the pranker seems to know most people are watching this out of disgust for him rather than any appreciation for the videos he’s making. Despite what the internet will tell you, empathy is not dead, and watchers don’t picture themselves filming, to say the least.
Consider The Source
Is it illegal to lie on the internet? Well, no. But generally speaking, saying “my uncle who works at Nintendo says they’re going to add another dog to the pocket monster games” is not the same as “I work at Nintendo and we’re planning another dog for the games” and that is not the same as “We here at Nintendo, on our official Facebook page, are announcing a dog for the next pocket monster game”. People online lie all the time, for good and for bad, and the easiest way to avoid being fooled into doing something stupid or wrong is to consider the person who’s delivering it to you.
Run through a list: say you’re online, and you see a recipe for a cake that looks good.
Step 1: where are you seeing it? Is this a baker’s personal blog? Or Reddit, or TikTok? How anonymous is the person who posted it? Reddit has a terrible problem with creative writers not outright admitting they were writing a fictitious story instead of describing a real thing that happened to them. This was usually harmless enough… until the writer is trying to portray a stereotype about a particular group of people as though it really happened. They get their imaginary internet points, and bigots get their internal beliefs re-validated. It stopped happening as often when Redditors started digging through comment histories to force these people to admit they lied, but it’s still a problem. Made up patterns and recipes are a less harmful version of this phenomenon, where they’re still seeking engagement and internet points and don’t really care if you waste food and time trying to make what they made. At worst, they can always cut and run. On the flip side, a baker’s personal blog needs a recipe to be as good as possible, because there’s no simply ditching the account if a post flops unless the baker is willing to drop literally all of their branding to do so. They have a vested interest in ensuring the recipe is at least mediocre, so they don’t lose their customers’ good will.
Step 2: does it align with previously understood information? Most amateurs know that angel food cake is a different texture than pound cake, but they may not know why. An especially good recipe may have a foreword explaining the reasons some step is necessary, but a bad one will seem esoteric. I’m reminded of the “dalgona coffee” that was trending on TikTok in 2020, where the instructions were to “stir the ingredients 200 times” instead of, I don’t know, telling people to whip it until it was frothy? Telling them to use a milk frother if they had it? Stirring with a spoon 200 times was weird. It wasn’t AI in this case, but it was bait to make a recipe go viral.
Consider your own internal biases, too. If you think you’re an expert in something, you should be able to use that information to verify a claim outside of a knee-jerk “no, that sounds wrong”. For instance, a “magic cake” is a type of cake that separates into layers while baking, a layer of custard and a layer of cake. If you tried to make this recipe a ‘thing’ today, you’d get a lot of people saying “this seems fake” unless you gave a looong rundown on how it works beforehand, which is not always possible in short-form content. However, if you can find other recipes, or recipes from trustworthy places like America’s Test Kitchen that seem to parallel the recipe you’re reading, well – it might be real!
Step 3: How is it framed? Does the source want you to panic, or ‘like’ their video, or comment on something in the background? Does it seem especially annoying, or overly sappy? Before the last two and a half decades of 24 hour news, there was an understanding that news shouldn’t be inflammatory or designed to frighten people, it should be news. But, with the rise of forever-on networks that need people to keep tuning in to them instead of the other guys or the internet, everything has to tie back to the news station’s core conceit. We’re in a culture war and this story proves it, or we’re losing a battle and this story proves it are common scripts to tie local stories to national storylines. With the rise of the internet, it didn’t even really need to be true anymore, because lying on the internet is basically nothing. ID an account as run by a liar who has a grudge against Swedes for some reason, and that person can just deactivate, and start a new account, saying the same thing. You’d have to ID them again, and all they have to do is block you on their new account before you do that so you don’t see them and can’t call them out.
This Itself Isn’t a Fix
Considering the source is important, yes. But, in the event someone lies about a source to make their misinformation seem more legit, the other steps – considering the actual content, considering the way it’s framed, et cetera – might help you avoid making a kneejerk reaction like, say, selling all of your gold because you saw a video “published by the White House” but posted on TikTok by some account with a bunch of random letters and numbers in it’s account name.
The truth is that we aren’t ready for the tech that’s coming. It’s already being misused constantly, and it wasn’t like it had to be great to make ragebait. Online verification of identity is it’s own complete mess, but maybe it’s a good idea for organizations (like official government channels) to verify themselves in places where verification means something to ensure there’s at least one place where they are, for certain, the source, including new channels like TikTok where misinformation spreads even easier than wildfire.
