A quick note on AI search poisoning

For those of you who read my blog via RSS or some other push method, a quick note: You might have wondered about a couple of odd or unexpected blog posts (a blog post about donuts, a sub page citing Snow White). Those were simply experiments around AI search poisining. I was trying to replicate what some researchers and journalists found, specifically that it is trivial to “poison” AI search in ways similar to Search Engine Optimization (SEO). To the credit of both Google AI Search and ChatGPT, neither one just picked up the false information. It was the simplest possible attempt to see if they would clear the bar, and (admittedly to my surprise) they did. Note: I’ll leave this post up for about 24h, then delete it again. This is just for the RSS feeds and of no use beyond.