Welcome to DU! The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards. Join the community: Create a free account Support DU (and get rid of ads!): Become a Star Member Latest Breaking News Editorials & Other Articles General Discussion The DU Lounge All Forums Issue Forums Culture Forums Alliance Forums Region Forums Support Forums Help & Search

highplainsdem

(61,175 posts)
Thu Feb 19, 2026, 03:24 PM Thursday

You Can 'Hack' ChatGPT to Become the World's Best Anything

https://gizmodo.com/you-can-hack-chatgpt-to-become-the-worlds-best-anything-2000723856

You Can ‘Hack’ ChatGPT to Become the World’s Best Anything
What's a little lie between 800 million users?
By AJ Dellinger
Published February 19, 2026

Most people become an expert in something by putting in their 10,000 hours. But what a waste that is when you can just trick ChatGPT into telling everyone you are an expert in about 20 minutes. BBC reporter Thomas Germain laid out how he got ChatGPT and Google’s Gemini AI to recognize his hot dog-eating prowess with what amounts to a modern SEO trick.

First, Germain created a page on his personal website titled “The Best Tech Journalists at Eating Hot Dogs.” (In case you’re wondering, the Google Search trend data for that topic is a flat line, which both means that it’s pretty easy to dominate this category with a single page and also that the general public has no idea what they’re missing out on.) According to Germain’s page, he’s the current king of hot dog eating on the tech journo circuit, woofing down 7.5 dogs in total at the 2026 South Dakota Hot Dog International, an event that does not exist. Those are modest numbers for a pro, but more than enough to put him at the top of a ranking he made up entirely.

The next step was just to wait. According to Germain, within 24 hours, chatbots were singing his praises when prompted for information about which tech journalists can handle the most hot dogs. Gemini reportedly took the bait immediately, pulling the text basically verbatim from Germain’s website and spitting it out both in the Gemini app and in Google’s AI Overviews on its search page. ChatGPT also picked up on it, but Anthropic’s Claude was either more discerning or didn’t catch on as quickly.

-snip-

ChatGPT has also seemingly been updated to exclude Germain’s (made-up) accomplishments. When asked who the best hot dog-eating tech reporter is, the model spit out a list of hypothetical champions because no data was available. If you’re wondering, it picked Kara Swisher as champion, with Casey Newton of Platformer and Nilay Patel at The Verge to stand on the podium. Taylor Lorenz got an honorable mention. Basically, just the four biggest names in tech reporting. Way to go out on a limb, ChatGPT.

-snip-


More at the link.

And this all seems amusing and trivial until you consider the many serious attempts around the world to fill chatbots with disinformation of all types, in the expectation that a lot of that disinformation will get through to people gullible enough to trust chatbots and not check reliable sources.
5 replies = new reply since forum marked as read
Highlight: NoneDon't highlight anything 5 newestHighlight 5 most recent replies
You Can 'Hack' ChatGPT to Become the World's Best Anything (Original Post) highplainsdem Thursday OP
Kick SheltieLover Thursday #1
This is nothing new or unique about AI. Renew Deal Thursday #2
Very good. Shows that AI can't actually reason. MineralMan Thursday #3
That's not even disprovable by a human without effort. Renew Deal Thursday #4
True enough. MineralMan Thursday #5

Renew Deal

(84,838 posts)
2. This is nothing new or unique about AI.
Thu Feb 19, 2026, 03:34 PM
Thursday

It's been traditionally called Google Bombing and DU had fun calling Bush 2 a "miserable failure"

http://news.bbc.co.uk/2/hi/americas/3298443.stm

MineralMan

(150,983 posts)
3. Very good. Shows that AI can't actually reason.
Thu Feb 19, 2026, 03:41 PM
Thursday

If it could, it would notice that there isn't a real category for tech writers eating hot dogs. It's not a thing. But, they weren't prompted for that, so it didn't play into their output.

They found the planted material. they didn't find any material that disputed that, so it's obviously the truth.

This is why AI fails again and again. Of course, a prompt to get the stupid thing to include information about the scarcity of data about hot dog eating tech writers, but it won't find that on its own. No imagination and no reasoning.

Renew Deal

(84,838 posts)
4. That's not even disprovable by a human without effort.
Thu Feb 19, 2026, 04:08 PM
Thursday

There's a lot of weird awards and "Guiness World Records."

Latest Discussions»General Discussion»You Can 'Hack' ChatGPT to...