You open social media and see a video of a celebrity saying something completely ridiculous. The voice sounds real, the face moves naturally and the lighting looks convincing. So you watch it again.
But after a while, you stop in your tracts to question if it is indeed real or is it AI. Is it possible that an actual video was edited, or did the person genuinely say something that will require an apology by lunchtime? Well, it seems we have a problem and it ain’t going away anytime soon.
For all those who believe seeing is believing, unfortunately that doesn’t hold water any longer. Hearing is not particularly reliable either, and the person replying to your comment may not be a person at all. The internet has always contained lies, edited photographs and people pretending to be someone else. But artificial intelligence has made the process faster, cheaper and far more convincing.
The truth is that every photograph has turned into a small investigation. We automatically look for signs to authenticate the image before us. Unfortunately, even those clues are disappearing as the technology improves. AI-generated people no longer need six fingers or teeth that look like they have been arranged during an earthquake. And you know what else is disturbing? It is when the fake person looks more trustworthy than the real one.
Research from Lancaster University found that highly realistic AI-generated faces can fool people into believing they are genuine. This is not surprising. What is surprising is that some participants considered synthetic faces more trustworthy than photographs of actual humans. Yes, we have certainly created a strange place. My reasoning is simple. A person who never existed may now appear more dependable than someone who woke up, got dressed and stood in front of a camera. And images are only the beginning of this dilemma. AI can easily copy voices, create videos and place public figures inside situations that never happened. It can make someone appear to confess to a crime, engage in a sexual act or deliver a political speech from a room they have never entered. Have you considered that person could very well be you?
The International AI Safety Report 2026 describes cases in which criminals used AI-generated voices and identities to impersonate executives or family members. Some victims transferred money because the person on the call sounded exactly like someone they trusted. Now ask yourself, if you received
a desperate call from your child, spouse or a friend. Would you pause to ask whether an algorithm created the panic, or would you respond like a human being? I am willing to bet that as long as the voice was familiar and they addressed you by your first name, that’s all that matters. That is what makes this technology so effective. It does not only imitate appearances but also uses trust against us.
Something we hardly ever question is a comment. Open almost any popular post and you will find hundreds of replies agreeing, arguing and asking questions. But have you stopped to consider how many came from actual people? I’m sure you are aware that bots can now produce responses that sound casual, emotional and reasonably informed. Isn’t it worrying that they can support a political position and make an unpopular opinion appear widely accepted? You see, all it takes is for enough artificial accounts to repeat the same message. And almost instantly, a lie begins to look like public opinion.
For me, this completely changes the way we understand online conversations. Because I once assumed that a comment represented a person, despite the terrible spelling and an extremely questionable profile picture. Now I question whether the conversation itself exists. Perhaps two humans are arguing. Or perhaps 500 bots are creating the appearance of an argument while the humans watch. How wonderful.
A 2026 Malwarebytes report found that 74% of respondents worried about encountering deepfakes or other AI-generated scams. That concern not only makes people suspicious of false content. It may also make them doubt genuine evidence, which could become the more dangerous problem. It may also make them doubt genuine evidence which could become the more dangerous problem. Think about it, if a real video shows someone behaving badly, they can simply claim that AI created it. If an authentic recording exposes corruption or abuse, supporters can dismiss it as a deepfake before anyone has examined the evidence. This proves that AI gives liars two useful tools. First it allows them to create something false. Second it gives them an excuse to reject something true.
So, what exactly can we trust?
Technology companies are developing systems called Content Credentials, which can record information about where an image came from and how someone edited it. The US Office of Research Integrity describes these credentials as one way to verify the history of digital material. Now that may help, but I hardly think that it will solve everything.
A missing label does not automatically mean an image is genuine. Metadata can disappear when someone takes a screenshot, downloads a file or moves it between platforms. And most people will not perform a technical inspection before watching a funny video during lunch. Platforms therefore need to carry more responsibility. Clear labels should not be hidden behind menus. Companies should have controls in place that identify automated accounts and react quickly when someone uses a real person’s face or voice without permission.
But in saying that, we also need to change our own behaviour.
Before sharing something shocking, stop for a moment and check where it came from. Look for the original report and see whether reliable sources have confirmed it. If a family member suddenly asks for money, contact them through another method before selling your house to help them. Basically, we must all become part-time investigators because the internet has decided that simply enjoying ourselves was too easy. Now I’m not saying that AI itself is the villain in every situation. AI can help people create, communicate and solve problems. And it can also be used for harmless entertainment, visual effects and ideas that would otherwise remain inside someone’s imagination.
The issue is not that synthetic content exists. The issue is that it can enter the same spaces as journalism, evidence and human conversation without clearly identifying itself. Trust once depended on what we could see and hear. Now it requires context, verification and a willingness to pause before believing the most exciting version of events.
We are entering a world where almost anything can be created. The question is whether we are also creating a world where nobody believes anything.
