You can tell clients "LLMs are probabilistic" until you're blue in the face, but I don't think most people fully appreciate how much of a mind bender it is for GTM teams who want to win AEO but don't live and breathe this stuff. We all started our careers working with systems where the relationship between inputs and outputs was predictable, but LLMs are not. With LLMs, a brand might show up consistently one week and disappear the next, even if nothing meaningful has changed about the brand or its marketing. And trying to figure out why less drastic, more subtle changes happen is typically a mix of guesses and bias. All this makes it difficult to distinguish between what's working, what's just noise — and when you are successful, what to double down on. We've all been trained to look for patterns, establish benchmarks, build repeatable processes. so naturally, we're trying to force AEO into the same box, but it isn't going to work. We need to get comfortable with the idea that we can't engineer certainty or predictability into AEO programs. That doesn't mean we stop measuring, testing, or trying to improve outcomes. However, I think we, as an industry, need to rethink what success looks like, the expectations we're trying to set, and how much confidence we should have in the conclusions we're drawing.