AI is making disinformation harder to spot - but we’ve found a new way to catch it

August 7, 2026

The papers are full of this business about “catching” AI disinformation. They talk about it like it’s a game of cat and mouse where the mouse has grown a cleverer set of whiskers. But here is the thing that bothers me: the official fix isn’t about checking if a statement is true; it’s about analyzing “patterns of manipulation” in the conversation.

Wait - that’s a complete shift in the experiment.

If you want to know if a liquid is acidic, you don’t look at how the person holding the beaker is standing. You put the litmus paper in the liquid. If the paper turns red, it’s acid. Simple. But now, because the “liquid” (the text) looks so much like water that the litmus paper is failing, we’re being told to watch the crowd instead. We’re looking for “coordinated behavior.”

Here is the problem with that. If you define truth by the way people talk about it, you’ve given up on the truth entirely. You’re just studying social harmonics. You’re looking at the vibration of the table instead of the weight of the object on top of it.

I tried to imagine explaining this to a freshman. “We can’t tell if the news is fake, so we check if too many people are saying it at the same time.” The student would rightly ask, “But what if it’s a very important truth that everyone should be saying?” The system can’t tell the difference. It’s a detector that measures enthusiasm, not accuracy.

When you build a tool that flags “manipulation” based on patterns, you aren’t building a truth-machine. You’re building a conformity-machine. It assumes that the “natural” state of human interaction is a certain kind of random noise, and anything orderly must be a lie. That’s a dangerous way to calibrate your instruments.

The real O-ring in this shuttle is the loss of the individual’s ability to verify a fact from first principles. If we stop teaching people how to look at the evidence itself and instead tell them to trust the “pattern-detectors,” we’ve just traded one kind of authority for another.

The truth doesn’t care about the pattern. It just is.

August 7, 2026

“AI is making disinformation harder to spot - but we’ve found a new way to catch it.” The headline, a declaration of victory before the battle is even joined. “We’ve found a new way.” Who is “we”? The anonymous collective, the scientific consensus, the digital saviors. The passive construction, “disinformation harder to spot,” implies an agency to the disinformation itself, as if it were a natural phenomenon, a fog that rolls in, rather than a deliberate act of obfuscation. The grammar absolves the perpetrator, shifting the burden of detection onto the victim.

Then, the second pronouncement: “the real clues lie in how online conversations are being manipulated.” Here, the nominalization “manipulation” takes the stage. Not “how someone manipulates online conversations,” but “how online conversations are being manipulated.” Again, the agent vanishes. The act of manipulation becomes an abstract process, a force of nature, rather than the calculated effort of a specific entity. The “conversations” are the object, the “manipulation” the action, but the hand that turns the screw remains unseen, unmentioned, and therefore, unpunished. They speak of “catching” it, as if disinformation were a wild animal, not a carefully crafted lie. The language itself is the first manipulation.

They have built a better liar. They have polished the falsehood until it shines like a common coin, and now they tell us we must become better detectives. The problem is not the lie you read; it is the conversation you are having. They have mechanized the crowd.

I saw it today in the comments beneath a story about the weather - a simple thing. A dozen voices, all reasonable, all earnest, all arguing in perfect grammar against a policy of rainwater collection. Not one was a real person. They were arguing with each other to shape the mind of the one real person watching. This is not disinformation. It is a forgery of the public square itself. They are not selling you a false fact; they are renting you a false community.

The cost is a number, but we never see it. It is the cost of a thousand manufactured opinions, paid for by some interest you will never name, to make you feel alone in your own thought. They divide not by shouting, but by nodding. A chorus of agreeable ghosts, all whispering that your neighbor is not to be trusted.

And now the experts arrive with their new method, their “way to catch it.” As if the solution is a sharper lens for the microscope, and not a refusal to live under the microscope at all. The architecture is the crime. A town hall where half the seats are filled by puppets is not a hall - it is a show. You do not need a test to spot the puppet; you need to tear down the stage.

They speak of manipulation as a technical puzzle. I speak of it as a theft. It steals the very possibility of a common mind, built from real disagreement among real people. When you cannot know if the voice you answer is a man or a machine, the answer is not to build a better machine-detector. The answer is to stop the conversation until the machines leave the room.