Last year, I handed the impossible task of predicting March Madness to ChatGPT.
The result? Sadly, my bracket was as busted as everyone else’s, but at least I believed my picks had more substance than luck.
A lot has changed in 365 days, and this time around, I’ve assembled the AI equivalent of the Dream Team: Google’s Gemini Researcher, OpenAI’s GPT-4.5 Research Agent, xAI’s Grok-3 with Reasoning and Research, Baidu’s freshly released ERNIE X1—which is supposed to beat DeepSeek R1 and all of OpenAI’s reasoning models. I worked up a bracket for each of them, which I shared below.
Then I dusted off my old GPT-based bot—which is now supercharged with GPT 4.o under the hood—and updated it with the past season’s data.
Here it is in case you want to fool around with it yourself: ask it to prioritize a specific methodology or statistic, analyze each game separately, or tweak the way it ponders information so it fits your needs and style.
Maybe one of these AIs will do better than last year, but probably not: The odds of picking a perfect bracket remain ludicrous at 1 in 9.2 quintillion. If you’ve filled out brackets since the dawn of humanity, you’d still have a 99.9999999883% chance of getting it wrong.
But these AIs don’t care about the odds. They’ve crunched terabytes of data, analyzed every Ken…
