Remember when spotting a deepfake was as simple as counting fingers? A little too many knuckles, a blurry patch around the jaw, eyes that seemed to follow you in a way that wasn’t quite human. For a while, those visual glitches were our safety net. But that net has been pulled away. Modern AI-generated images and videos no longer come with the telltale weirdness that used to give them away, and the tiny flaws we learn to spot today are already being patched out of tomorrow’s models. We are left in a strange place: the internet can show us anything, and most of us can no longer tell what’s real. The consequences are not just embarrassing; they’re dangerous. Consider a few recent examples. Mitch McConnell’s hospital jeans were real, but the image of Nekima Levy Armstrong, an ICE protestor turned into a symbol of “far-left agitators” by the Trump administration, was faked. President Trump’s outing with Mike Tyson was legitimate, but the viral photo of an adulterous Georgia city mayor was not. Meanwhile, digital scams involving deepfakes caused more than $148 billion in losses in 2025, a 25 percent increase from the year before. If the old visual Turing Test no longer protects us, what should we do instead? The experts quoted in this piece offer a surprisingly human answer. Stop trying to inspect every pixel. Start paying attention to the things that make us human: our ability to reason, our awareness of our own emotional reactions, and our willingness to seek out trustworthy sources. It is not as satisfying as spotting a fake in a single glance, but it works—and it will continue to work long after today’s AI artifacts disappear. The shift may feel uncomfortable at first, because we’ve been trained to look for errors in the image itself, not in the story around it. But the more we practice this new kind of skepticism, the more natural it becomes, and the harder we are to fool.
The first and most important rule, according to cybersecurity experts, is to stop looking for glitches and start verifying sources. That advice applies especially to video, where generative models are improving so quickly that any flaw you learn to recognize today will likely be fixed by next month. “Every detection tip we publish eventually becomes a line in someone’s prompt,” explains Roei Sherman, senior director of research at cloud security firm Mitiga. In other words, the moment we teach people to look for a certain shadow or reflection, AI developers feed that observation into their next generation of tools. Instead of hunting for artifacts, Sherman suggests finding the larger context from which a clip or image came. If you see a short video of a political candidate saying something outrageous, look for the full speech or interview. Search for coverage of the same event from multiple reputable outlets. “Short clips cut out of context are where most of the manipulation lives,” he says. This is harder than it sounds, because even our search tools are becoming less reliable. Sofia Rubinson, senior editor at misinformation watchdog NewsGuard, points out that reverse image search on Google can still help, but Google’s new AI-powered search features sometimes repeat the very misinformation you’re trying to check. Fact-checkers used to rely on traditional search to trace the origins of photos and videos, she says, but now AI-generated summaries often misattribute images and reinforce the hoax. So when you trace something back to a source, don’t just trust that the source is legitimate because it looks polished. Ask practical questions: Has this account existed for more than a month? Does this information appear anywhere else, from anyone with a familiar, credible name attached? Is this person asking me to act right now, and if so, why? Marijus Briedis, chief technology officer at NordVPN, notes that none of these questions require you to analyze a pixel or a waveform. That is exactly why they still work. They rely on human judgment rather than technical knowledge, and they don’t become obsolete the moment AI gets better at rendering hands or shadows.
The second tool in your defense is simple logic. Experts say you should go with your gut and always consider plausibility. How likely is it that this event actually happened? If the post appears to come from a public figure, would that person really say such a thing on a public platform? If a story seems designed to confirm your worst suspicions about someone you already dislike, pause before sharing it. Rubinson stresses the importance of looking for multiple angles of an event or crisis before deciding it’s legitimate. “Keep an eye out for corroborating information,” she urges. This is especially important because a lot of social media is fake, and even the stories that feel true sometimes turn out to be staged. But the logic test doesn’t only apply to wildly unusual events, like a video of bunnies hopping on a trampoline or a politician morphing into a cartoon character. Some of the most effective misinformation is subtle and believable: a fabricated comment from a politician, a doctored screenshot of a celebrity’s social media post, a fake headline that confirms something you’ve suspected for years. Rubinson warns that incendiary comments from politicians and controversial actions from celebrities deserve to be double-checked, even if many of them turn out to be real. The point is not to assume everything is false; it’s to give your brain a moment to ask whether the story makes sense in the real world. If it feels too outrageous to be true, it might be exactly that. If it feels perfectly aligned with your existing beliefs, that’s another reason to slow down. We all want to believe things that flatter our worldview, and manipulators know this better than anyone.
Your emotions are also a warning system, and you should pay attention to them. Sherman says it’s essential to remain aware of how an image or video makes you feel. If you have a visceral response to a post—if it enrages you, frightens you, or stirs a desperate need to share it immediately—that may be a sign that it was designed to do exactly that. Rage is one of the most powerful tools in the misinformation playbook. When content makes you angry, your critical thinking switches off, and your thumb moves toward the share button before your brain has a chance to check the facts. This is sometimes called rage-baiting, and it’s everywhere. Rubinson points out that if you see one wild video appearing repeatedly in your feed, especially one that seems deliberately crafted to evoke an emotional response, there is reason to be suspicious. The algorithm is not showing it to you because it’s true; it’s showing it to you because it keeps you engaged. Once you recognize that, you can start to separate your feelings from the facts. That doesn’t mean you should become cold or indifferent to injustice; it means you should use your emotional reaction as a cue to investigate further rather than as proof that something is real. Ask yourself: Why is this content making me feel this way? Who benefits from my anger? What action does this post want me to take? If the answer is “share it immediately” or “donate right now” or “hate this person,” then you are probably being manipulated. The best response is often to close the tab, take a breath, and come back later with a clearer head. By monitoring your own reaction, you turn your vulnerability into a strength.
Of course, there are tools that can help, and you should use them—but with caution. If you’ve examined the post, checked its sources, and still aren’t sure, turn to the professionals who track viral deepfakes. NewsGuard publishes an AI monitor and a newsletter on trending deepfakes and misinformation. Cybersecurity company McAfee flags popular AI-powered scams on its AI news blog. Deepfake detector company AI or Not maintains a searchable database of known deepfakes and manipulated images. These resources can be valuable, especially when you’re trying to figure out whether something you saw is part of a larger pattern. There is also progress happening behind the scenes: while federal oversight lags, state-level regulations and industry-wide content provenance measures are slowly making it harder for fakes to spread undetected. Cybersecurity experts are building increasingly robust countermeasures, including more advanced AI deepfake detectors. But here is the uncomfortable truth: no detection tool is reliable enough to issue a definitive verdict on whether something is AI-generated. A tool might say an image is 90 percent likely to be fake, but that 10 percent chance still matters, especially when reputations are at stake. Briedis offers a helpful way to think about these tools. They are not judges; they are companions. “What they are good for is the moment something feels off,” he says. “The tool agrees with you, and now you go and verify it properly.” In other words, use a detector as a prompt for further investigation, not as a final answer. If the tool says something is fake but you can’t find corroborating evidence, keep looking. If the tool says something is real but your gut tells you otherwise, trust your gut enough to dig deeper. The goal is not to find a machine that can think for you; it’s to find a machine that can help you think better.
In the end, the most practical advice is also the simplest: if you’re still not certain an image or video is real, don’t share it. That may sound overly cautious, but it’s the only way to stop the spread of misinformation in a world where fakes are becoming indistinguishable from reality. Every share gives false content more life, more reach, and more power. Before you pass something along, ask yourself whether you would be comfortable explaining to someone why you posted it. If the answer is no, let it go. This is not about being paranoid or refusing to believe anything you see online; it’s about being humble enough to admit that you might be fooled. The people who create deepfakes and AI hoaxes are counting on our confidence. They know that most of us won’t bother to check the source, that we’ll trust our emotions, and that we’ll share things in the heat of the moment. By doing the opposite—by being skeptical, patient, and willing to say “I don’t know”—you become a much harder target. You also become a better citizen of the internet, someone who helps slow the spread of lies rather than accelerating it. The experts in this article all agree on one thing: the visual tells are gone, and they’re not coming back. But that doesn’t mean we are helpless. We still have our brains, our instincts, and our ability to connect with reliable sources of information. We still have each other. The next time you see a shocking image, a wild video, or a suspicious claim, take a breath. Look for the source. Ask if it makes sense. Notice how it makes you feel. Use the tools available to you, but don’t let them make the decision for you. And if you’re still unsure, resist the urge to share. In a world of endless fakes, the most human thing you can do is hesitate.