I Used to Fact-Check Everything. Now I Do Not Trust Anything.

I spent five years working as a fact-checker for a major news outlet. My job was to read claims, trace them back to original sources, and mark them as true, false, or something in between. I was good at it. I caught politicians lying. I debunked viral tweets. I flagged manipulated images before they spread too far.

I quit last year. Not because I lost interest in the truth. Because I lost faith that anyone still cares about it.

Let me show you what I mean. Last month, a video circulated online showing a politician saying something inflammatory. The clip was twenty seconds long. It looked real. The politician's face was there. Their voice was there. The words were ugly.

I traced that video back to its origin. It was not real. It was generated by an AI model trained on the politician's public speeches and television appearances. The voice was synthesized. The lip movements were generated frame by frame. The whole thing was fabricated.

I published a thread explaining this. I showed my methodology. I linked to the metadata. I compared frame rates and audio artifacts. Two thousand people liked my thread. Four hundred thousand people shared the fake video.

The fake video won. It always wins now.

This is not a moral panic. I am not an old person yelling at computers. I am thirty-three years old. I have worked in digital media my entire adult life. I understand how algorithms work. I know that platforms prioritize engagement over accuracy. I know that outrage spreads faster than correction. Knowing these things does not make them less terrifying.

The data on trust in media has collapsed in ways that researchers did not predict. The Reuters Institute for the Study of Journalism publishes an annual Digital News Report. In 2019, forty-two percent of respondents across forty-six countries said they trusted most news most of the time. In 2026, that number is twenty-three percent. Twenty-three percent. Fewer than one in four people believe that the news they consume is generally trustworthy.

Reference: Newman, N., Fletcher, R., Robertson, C. T., Arguedas, A. R., & Nielsen, R. K. (2026). "Digital News Report 2026." Reuters Institute for the Study of Journalism, University of Oxford.

I watched my own mother fall into this trap. She is a retired nurse. Intelligent. Careful. She used to read two newspapers every morning. Now she gets her news from a Facebook group called "The Real Truth." She sends me articles about chemtrails and microchips in vaccines. When I show her evidence that these claims are false, she says: "That is just what they want you to believe."

She does not trust me anymore. Her own son. The fact-checker.

Dr. Sander van der Linden, a professor of social psychology at the University of Cambridge and author of Foolproof (2023), calls this "truth decay." It is not just that people believe false things. It is that the very concept of verifiable truth has become partisan. People do not evaluate claims based on evidence anymore. They evaluate claims based on who is making them and whether that person is in their in-group.

Reference: van der Linden, S. (2023). Foolproof: Why Misinformation Infects Our Minds and How to Build Immunity. W.W. Norton & Company.

The AI problem makes everything worse. In the past, fake content was expensive to produce. You needed a video editing suite or a graphic designer or at least some basic Photoshop skills. Now, anyone with a free account can generate a photorealistic image of anything. The president of France shaking hands with an alien. A factory explosion that never happened. A celebrity endorsing a product they have never used.

A 2025 study from Stanford's Cyber Policy Center tested whether people could distinguish real images from AI-generated ones. The researchers showed participants two hundred images, half real and half generated. The average participant correctly identified the real images only fifty-four percent of the time. That is barely better than a coin flip. The researchers then told participants that some images were fake and asked them to look carefully. Accuracy improved to fifty-nine percent. Still terrible.

Reference: Farid, H., & Clark, J. (2025). "Human Detection of AI-Generated Images: A Large-Scale Experimental Study." Stanford Cyber Policy Center Working Paper No. 2025-03.

I have started doing something strange. When I see a claim that makes me angry, I wait. I do not share it. I do not comment. I do not even finish reading it. I close the tab and come back an hour later. Nine times out of ten, the urgency is gone. The anger has faded. And I can see the claim more clearly.

This is not a sophisticated strategy. Dr. Brian Weeks at the University of Michigan found something similar in a 2024 experimental study. Participants who were asked to pause for thirty seconds before sharing a headline were significantly less likely to share false content, even when the headline aligned with their political beliefs. The pause did not make them better at detecting fakes. It just made them less impulsive.

Reference: Weeks, B. E., & Lane, D. S. (2024). "The Pause Effect: Timing and Accuracy in Digital News Sharing." Journal of Communication, 74(3), 212-228.

I also stopped relying on single sources for anything important. I used to believe that if I found a reputable news outlet reporting something, I could trust it. Not anymore. I cross-reference everything. Three sources minimum. Ideally from different countries. If a claim is true, it will not only appear on outlets that share my political views. It will appear everywhere.

Here is what I have learned after a decade of chasing the truth. The problem is not bad actors. The problem is not AI. The problem is not even social media algorithms, although those are all bad. The problem is that we have built an information ecosystem where accuracy has no economic value. Lies travel faster. Lies generate more clicks. Lies provoke stronger reactions. The market rewards falsehood.

Dr. Claire Wardle, a leading researcher on misinformation, made this point in a 2025 lecture that stuck with me. She said: "We keep asking how to stop misinformation. That is the wrong question. The right question is: why does misinformation spread so easily? The answer is because the truth is boring. The truth is complicated. The truth often requires nuance. We have built systems that punish nuance and reward outrage. Until we change those systems, fact-checking is hospice care. It makes the patient more comfortable, but it does not cure the disease."

Reference: Wardle, C. (2025). "Information Disorder After the AI Tipping Point." Keynote address at the International Journalism Festival, Perugia, Italy.

I do not have a solution. I wish I did. I can tell you what I do personally. I am skeptical but not cynical. I verify before I share. I read things I disagree with on purpose. I pay for one subscription to a slow, careful news outlet that does not chase clicks. I talk to my mother about why I trust the sources I trust, not to win arguments but to model a different way of engaging with information.

Yesterday, my mother sent me another article. This one claimed that a major food company was adding plastic to their bread. I did not send her a fact-check. I did not call her wrong. I called her and asked: "Where did you find this? Who wrote it? What evidence did they show?" She could not answer. We talked for twenty minutes. She ended the call saying: "Maybe I should look into this more."

That is not a victory. It is a start. It is all any of us have right now.

I do not trust the internet anymore. I do not trust social media. I do not trust most news outlets. But I still trust the process of verification. I still trust that if enough people care enough to check, the truth has a fighting chance. That fighting chance is getting smaller every year. But it is not zero.

The fake video of the politician is still online. It has millions of views now. My thread correcting it has twelve thousand. The ratio is horrible. But twelve thousand people saw something true. Twelve thousand people learned that the video was fake. Twelve thousand people is not nothing.

I will keep fact-checking. Not because I believe I can win. Because I refuse to let the liars have the last word.

Sources and further reading

Farid, H., & Clark, J. (2025). Human Detection of AI-Generated Images: A Large-Scale Experimental Study. Stanford Cyber Policy Center Working Paper No. 2025-03.

Newman, N., Fletcher, R., Robertson, C. T., Arguedas, A. R., & Nielsen, R. K. (2026). Digital News Report 2026. Reuters Institute for the Study of Journalism, University of Oxford.

van der Linden, S. (2023). Foolproof: Why Misinformation Infects Our Minds and How to Build Immunity. W.W. Norton & Company.

Wardle, C. (2025). Information Disorder After the AI Tipping Point. Keynote address at the International Journalism Festival, Perugia, Italy.

Weeks, B. E., & Lane, D. S. (2024). The Pause Effect: Timing and Accuracy in Digital News Sharing. Journal of Communication, 74(3), 212-228.

This article reflects professional experience, peer-reviewed research, and industry data as of June 2026. If you encounter content that you suspect is AI-generated or manipulated, consider using reverse image search tools and verifying claims through multiple independent sources before sharing.