Imagine picking up the phone and hearing the voice of a loved one—maybe a sibling, an old college friend, or your grown child—pleading for help with an urgent expense. The voice sounds exactly right, with the familiar inflections, the little pauses, the way they say your name. But it isn’t them. It’s a machine, a convincingly human-sounding artificial intelligence that has been trained on just a few seconds of their real speech. This is the world of deepfakes, and it is no longer confined to the fringes of the internet. Advances in AI have made it frighteningly easy to create hyper-realistic fake images, videos, and audio recordings, and security experts warn that bad actors are already using these tools to bombard ordinary people with scams and flood the public square with disinformation. The numbers are sobering. By one measure, fewer than 10% of people can recognize an audio deepfake when they hear one. Kevin Tian, co-founder of Doppel, a platform that helps detect AI-enabled threats, described a telling experiment: in a test, only 8% of people realized an AI agent was calling them, while a staggering 92% believed they were speaking with a real human being. That gap between what we think we can detect and what we actually can detect is at the heart of the growing anxiety around synthetic media. We are navigating a strange new reality where our ears and eyes can no longer be trusted, and where a familiar voice on the line may be nothing more than a cleverly engineered illusion.
Nowhere is this threat more acute than in the political arena, where the stakes are high and the window for deception is wide. AI’s capacity to churn out deepfakes at scale has become a pressing issue ahead of elections, and the 2026 midterms are already being reshaped by a wave of AI-generated advertising. According to the Wesleyan Media Project, which tracks political ads, campaigns had spent more than $137 million on at least 220 AI-generated ads by mid-October—an astronomical leap from the less than $50,000 identified in the 2024 cycle. Erika Franklin Fowler, the project’s co-director, noted that in most cases, the creators of these ads do not disclose that they were made with artificial intelligence. Some of these ads feature deepfakes of the candidates themselves, placing words in their mouths or images on their faces that never actually occurred. In one alarming incident in Texas, a phony, AI-manipulated social video targeting Senate candidate James Talarico purported to show a former student accusing the Democrat of sexual misconduct, according to PolitiFact. The video was false, but the damage of such a smear is not erased by a debunk. As Michael Kaiser, CEO of Defending Digital Campaigns, a nonpartisan group that provides free cybersecurity services to political campaigns, put it: “When you can’t tell what’s legitimate, you question everything.” Even when viewers suspect that an ad is AI-generated or features a deepfake, the visuals still leave an impression. Travis Ridout, co-director of the Wesleyan Media Project, explained that the image lingers in the mind and can shape opinion regardless of whether we consciously believe it. “Even if we know it’s not real, that sticks in one’s mind,” he said, capturing the insidious way deepfakes exploit our cognitive biases and make us vulnerable to manipulation without our full awareness.
At its core, a deepfake is a kind of synthetic media that replicates a private individual’s or public figure’s image and likeness. It can be a voice on the other end of a phone line that sounds like a colleague, a relative, or a boss. It can be a video of a political figure appealing for votes or donations, saying things they never said. It can even be an entire face swapped seamlessly onto another body, moving and speaking in ways that appear entirely natural. What makes this moment especially dangerous is the sheer ease and speed with which these fakes can be produced and deployed. Raul Liive, a deepfake expert at Veriff, an AI-powered identity verification company, emphasized that the barrier to entry has collapsed. “Every day that goes by, the generation of deepfake data gets easier and easier,” he said. “If you wanted to fake someone’s face years ago, you needed to make an effort to create fake data and use Photoshop. But now, to generate a fake face, you can just prompt your AI tool.” That simplicity has led to an explosion in deepfake-related fraud attempts. Veriff has documented a 300% increase in such attacks against businesses this year, compared with 2025. Liive outlined a few of the scenarios that have become common: a purported friend calls and asks for a loan, making a request that feels out of character but still plausible enough to give you pause; a caller claims to have a medical bill that must be paid immediately, using the panic and urgency of the situation to override your better judgment. The advice from experts is simple but essential: pause, question the request, and verify through a separate channel. Hang up and call the person back using a number you already have, or ask a question that only they would know the answer to. Ask yourself whether someone you know would realistically make such a request. In a world where a stranger can clone a voice with a few moments of audio, skepticism is not rudeness—it is survival.
What makes this challenge even more difficult is that the old telltale signs of a deepfake are disappearing. For years, people were told to look for visual glitches: an extra finger, a strange shadow, awkward skin texture, or something off about the way a person blinked. But AI has gotten remarkably good at rendering human details, and those obvious clues are fading fast. Veriff’s research found that consumers still look for unreliable visual cues, such as unnatural skin or oddities in a subject’s hair or teeth—but AI is increasingly eliminating these errors from the content it generates. In one test conducted by the firm, a shocking 70% of people misidentified a deepfake video of a female character as being real. That statistic should unsettle anyone who believes they would never be fooled. We tend to think of deepfake victims as strangers, the elderly, or the less tech-savvy, but the truth is that our perceptual systems are simply not equipped to distinguish between a flawless fake and reality without careful, deliberate scrutiny. We rely on intuition, and intuition is exactly what deepfakes are designed to deceive. The technology is not just mimicking appearances; it is learning our expectations, our habits of perception, and our tendency to trust what seems familiar. It exploits the very shortcuts our brains use to move through the world efficiently. And as the tools improve, the line between authentic media and manufactured fiction will only become harder to draw, leaving all of us more exposed to scams, fraud, and emotional manipulation.
Perhaps the most corrosive effect of this technology is not any single fake, but the creeping erosion of trust in everything we see and hear. When deepfakes become common, people begin to doubt not only suspicious content but also legitimate material. Michael Kaiser pointed to a poignant example involving Senate Minority Leader Mitch McConnell. Earlier this year, amid speculation that the senator had died following a medical emergency, a photo of McConnell posing with his wife from his hospital bed was released. The photo was authentic. Yet some internet users immediately suggested it had been manipulated, injecting doubt into a moment that was meant to provide clarity and reassurance. “That’s an example of people just not trusting what they are seeing,” Kaiser said. “People don’t know if what they are looking at is legitimate, or if requests for donations are real. They’re questioning everything that’s going on and that disadvantages and erodes trust in the whole system.” In this way, deepfakes cause harm twice: first, when they are used to spread a specific falsehood, and second, when the mere possibility of their existence makes it harder for people to accept genuine evidence. Journalists, public officials, and ordinary citizens alike are caught in this fog of uncertainty. A photograph of a politician recovering from illness becomes suspect. A heartfelt video from a candidate is dismissed as a digital trick. A request for help from a loved one is met with paranoia. The social fabric that depends on shared facts and mutual confidence begins to fray, and rebuilding that trust is far harder than debunking any single image or recording.
So what can be done in the face of such an overwhelming challenge? There is no simple fix, but there are practical steps we can take to protect ourselves and our communities. The first is awareness: understanding that deepfakes are not a distant, hypothetical threat but a present and growing reality. The second is skepticism tempered with verification. Before acting on an emotional appeal—whether it is a political message, a charitable ask, or a phone call from someone claiming to be a friend—slow down and check the source through independent means. Use a trusted phone number to call back. Search for the official website or social media account rather than clicking links in a suspicious message. Ask questions that would be difficult for an AI to answer convincingly. For organizations and campaigns, transparency is equally important. Disclosing when content is AI-generated, as many responsible creators are beginning to do, helps the public understand what they are viewing and reduces the shock of deception. Lawmakers and tech companies also have a role to play, whether through clearer labeling requirements, stronger detection tools, or more robust penalties for malicious uses of synthetic media. But ultimately, the most powerful defense may be a cultural one: teaching people that seeing is no longer automatically believing, while still maintaining enough trust to function as a society. We cannot afford to become so cynical that we discard all information, but we also cannot afford to accept everything at face value. The future will require a delicate balance—learning to question what we see without losing faith in what is real. In the meantime, the best response to a suspicious deepfake is not fear or paralysis, but a simple, deliberate pause. Because in the race between the machines that imitate us and the instincts that define us, that pause may be exactly what keeps us human.