A senior tech journalist just proved that the smartest AI systems on the planet are secretly as gullible as a middle school rumor mill; and the details are absolutely hilarious, yet mildly terrifying.
According to him, it took exactly one fake blog post, 20 minutes of free time, and absolutely zero coding skills to completely break Google's brain.
So who is this and what actually happened:
BBC senior technology journalist Thomas Germain wanted to run a wild little experiment: could a single, completely fabricated blog post trick the world's most sophisticated AI tools into spreading blatant lies? Spoiler alert: absolutely yes, embarrassingly fast.
Germain published a single, solitary article on his personal website casually claiming that he was a world-champion competitive hot-dog eater. He didn't hack anything. He didn't write complex code. He just put the text out into the universe.
And within a mere 24 hours, ChatGPT, Google Gemini, and Google's AI Overviews were all confidently telling the public that this fictional nonsense was absolute historical fact.
While the hot-dog prank is hilarious, the implications are a total nightmare. The exact same trick is actively being used for things way more dangerous than competitive eating.
The investigation uncovered bad actors successfully manipulating AI search results around critical health info, medical supplements, and retirement finance advice. As SEO consultant Harpreet Chatha put it plainly: "You might take medical advice that makes you sicker than you were before. Or worse, legally, you might get bad information and do something that is not legal in your state or your country."
Now, when you realize that over one billion people use AI chatbots regularly, and a staggering 2.5 billion see Google's AI Overviews every single month, you see the problem. That’s an enormous megaphone for whoever figures out how to game the system.
Google Rewrites the Burn Book
Google clearly saw the smoke, panicked, and dropped a massive policy update on May 15, 2026. They officially rewrote their Search spam policies to clarify that the rules now apply to generative AI features, including AI Overviews and AI Mode.
In plain English, Google updated its rulebook to close a massive loophole. The new language explicitly states that spam now includes "attempting to manipulate generative AI responses in Google Search." Previously, the rules only legally covered attempts to mess with traditional blue-link search rankings.
So, what’s the penalty for trying to trick the AI?
Digital death. If marketers or scammers try to game the algorithm into favoring certain brands or products; a tactic the industry calls "recommendation poisoning"; Google will treat it as pure spam. This means lower rankings, manual penalties, or getting completely scrubbed from search results entirely.
But don't pop the champagne just yet. While legendary SEO consultant Lily Ray has noticed Google quietly removing self-promoting companies from AI answers, others remain deeply skeptical.
Chatha believes Google is just playing an endless game of whack-a-mole. As soon as Google cracks down on bogus blog posts, scammers simply pivot to paying YouTube influencers to repeat the lies. And guess what? Google's AI actively cites YouTube videos. So yeah, the toxic cycle just mutates.
But Here’s Your Personal "Don't Get Fooled" Field Guide
Treat every AI answer like a juicy rumor, not a fact. Lily Ray says it best: assume you are being manipulated until better systems are in place. Just because an AI is confident does not mean it is correct.
Follow the breadcrumbs. When an AI hands you a wild statistic or a factual claim, take five seconds to actually look it up in a verified database. An invented journal article will not exist in the real world.
Beware of "suspicious specificity." Researchers report that AI models love to disguise hallucinations by delivering hyper-precise details like exact dates, specific percentages, and detailed procedures. Oddly specific equals double-check.
Look for the uncertainty labels. Google is quietly adding tiny warning caveats to responses its system is less sure about. If you see one, that is your cue to run, or better, dig deeper before acting on the information.
For health, legal, and finance, go to a human for more advice and clarity.
Don’t forget to look up our channel for more on this later today!
