I tested what a rafting customer might find in search and AI answers, documented gaps, refined our public pages, and checked again. I did not buy more ads. I worked to make the trips and guide training we actually offered easier to understand. People don’t only search anymore They ask. And the systems they ask already understand the industry they’re asking about. In ours, they know the common failure modes: safety concerns, inconsistent service, unreliable crews. If a business doesn’t clearly address those, it doesn’t get the benefit of the doubt. It gets grouped in with everyone else. Sometimes worse. I saw it when ChatGPT recommended another outfitter for a high-water trip after our guides had just finished training. That story made me ask what a new customer could find about our training. Interpretation follows reality That interpretation isn’t based only on what you say about yourself. It’s based on what’s actually true, and on how legible that truth is. If the operation doesn’t back up the claim, nothing compounds. You can publish endlessly and stay invisible. I kept testing what showed up and refining pages that explained how the operation runs, how guides train and why the experience differs. In the recorded metric, AI search appearances rose from around 30 a month to around 1,800. That is visibility, not bookings, and it does not isolate the effect of any one change. What the testing taught me An answer from one assistant is a clue, not a permanent ranking. I look at what a customer can find, compare it with the real business, make missing facts clear, and check again. That takes repeated tests and judgment, not more claims about being the best.