It Was Labelled Fake. It Still Went Viral.

Here is a small, perfect parable of where we are with AI and truth. A fake, AI-generated post about the pop group KATSEYE, dressed up to look exactly like an official announcement, went viral on X. The platform did its job and slapped a “Made with AI” label on it. And the views kept climbing anyway, past 2.7 million by mid-September. The safeguard worked flawlessly and changed nothing. Turns out a warning label is no match for a good-looking lie.

Why the label failed

Two reasons, both instructive. First, placement: on many platforms the AI label is not exactly in your face. It can sit under a “see more” tag or get crowded out by other information, so most people scrolling past never register it. A disclosure nobody reads is decoration, not protection. Second, and more fundamental: by the time you are told something is AI-generated, the image has already done its work. The fake borrowed the visual grammar of a real announcement, the official-looking layout, the confident formatting, and that format is what people trust at a glance. The label arrives after the impression has already landed.

There is a neat parallel here to how AI itself goes wrong. A hallucinated answer from a chatbot fools people for the same reason: it copies the format of a correct, authoritative answer, the citations, the confident tone, so it looks right even when it is nonsense. The KATSEYE hoax is that failure mode escaping the chatbot and loose on a social feed, wearing the costume of officialdom.

The takeaway

The comforting idea that we can label our way out of the AI-misinformation problem took a small, funny, telling knock here. Labels are worth having, but they are a speed bump, not a wall, and anyone counting on “Made with AI” tags to keep the internet honest is going to be disappointed. The durable defence is not a badge on the content, it is the habit in your own head: when something borrows the exact look of an official announcement, that polish is a reason to check, not to trust. The more real it looks, the more worth pausing on. These days, looking official is the cheapest thing in the world to fake.

Related: how AI is reshaping what we see and trust online.

Get the free weekly stack: the AI tools and moves that matter, hype filtered out.Subscribe free →
Scroll to Top