In 2023, about one in every 2,828 research papers had a fabricated citation. In the first seven weeks of this year, it was one in 277. That's not a rounding error. That's a trend line going straight up. And the scary part isn't that the citations are fake, it's that they're good. Real journals, real formatting, real-sounding co-authors, they just don't exist. Welcome back to the Dr. Verga podcast. It's been a while, but this is the place where we discuss all things wellness, wisdom, and warfare. I'm Jesse, and today we're talking about why it's backed by research has quietly become one of the least trustworthy sentences in the English language. Before we get tactical, let's get honest about why this matters. We've spent 20 years training people to treat a citation like a stamp of truth. Turns out that stamp is now something a chatbot can counterfeit on command, and most of us have never learned how to check it in the first place. The claim floating around every newsroom and university right now is that AI is quietly poisoning the well of human knowledge. Researchers auditing ARXIV, BioRXIV, SSRN, and PubMed found roughly 146,900 hallucinated citations sitting in papers on those repositories. At one recent AI conference, one in five sampled submissions had at least one fabricated reference, and reviewers, the actual peer reviewers, missed most of them. In July, a researcher signed an open letter asking universities to resist the hype and protect critical thinking expertise, academic freedom, and scientific integrity. That's not a headline from a tech blog. That's scientists asking other scientists to please stop trusting the robot. Here's why this works on smart people. Large language models are optimized to sound authoritative, not to be correct. When a model doesn't have a real source, it doesn't say I don't know. It generates something that looks exactly like a citation should look. Real journal name, plausible year, a real researcher's name attached to the paper, they never wrote. Your brain sees the format, author, year, journal, and does the rest of the work for it. Psychologists call this the authority heuristic. We outsource our judgment to the appearance of credibility instead of the substance underneath it. AI didn't invent that shortcut in your brain. It just learned how to exploit it at scale. And it's not confined to academia. FullFact's own investigation found Google's AI overviews repeating debunked claims, inventing details about real events, and failing to flag AI-generated images and video game footage as fake. Same mechanism, different venue. Something sounds official, so it gets treated as fact. And if you Google something but don't look past the AI overview, you're exactly who I'm talking to. This is exactly the gap my companion piece on W3 magazine introducing the W3 Evidence Index was built to close. A citation existing is not the same question as a citation deserving your trust. The Evidence Index scores sources 0 to 10 based on methodology, data quality, transparency, corroboration, bias risk, and who's paying for the report? A national probability-based survey and a random viral post with no named source are not equally backed by evidence, even though people cite both like they are. So here's the actual translation for you today. Before you repeat a claim because it cited a study, ask three questions. Can you find the source outside of the thing that cited it? Does the person or institution behind it have a reason to want you to believe this? And does the claim match what the source can actually support? Or is it doing more work than the evidence allows? Takes 15 seconds, saves you from being the person who confidently shares a hallucination at a dinner party. Are dinner parties still a thing? I think so. I want to be clear about what this isn't. This isn't trust nothing, and it isn't permission to dismiss real research because you didn't like the conclusion. Nihilism is just laziness wearing a skeptic costume. The goal isn't zero trust, it's calibrated trust. Some sources have earned a lot of confidence. Most claims deserve some. A few deserve almost none. The skill is telling the difference, not refusing to play the game. Speaking of telling the difference, this is exactly the muscle W3 magazine exists to build. Think deeper, live wiser, stay ready isn't a slogan I picked because it sounded nice on a business card. Every piece runs through actual evidentiary standards, including the evidence index framework I just walked you through. If you want the full companion article, the real greatest examples, including how a Pew survey scores against a vendor threat report, against a random viral post. It's live on W3 Magazine right now for free. And if you're not a subscriber yet, again, subscribing is free. Podcast listeners and watchers get a free 30-day trial by visiting the W3Magazine.com slash DoctorVirga podcast or the W3Mag.com slash DoctorVirga podcast or clicking the link in the description so I don't have to phonetically say that out for you guys. No cost for the first month, cancel whenever. And if it doesn't make you noticeably harder to fool, that's on me, not you. Links in the video description, back at it. So that's the citation problem. Think deeper about why a source existing was never the same as a source being trustworthy. Live wiser by running the 15 second check before you share anything, not after someone corrects you in the comments. And stay ready because the next version of this problem won't announce itself as a fake citation. It'll just look exactly like the real thing. If this helped, subscribe and send it to whoever in your group chat forwards the most study, show, text, new episodes every Monday. I'm Jesse. Stay ready.