I'm afraid most people who hear about GEO's generative engine optimization being exposed at the 315 Gala have a strange feeling. Is AI, this pure land, also going to be corroded by commercial speculation? The problem is actually not that simple: Because the core competitiveness of current generative engines, such as Perplexity or GPT with integrated search functionality, is "source aggregation". But the problem is that the model naturally carries the gene of "quantity worship" when crawling web data. In order to ensure the richness of answers, algorithms must lower the static filtering threshold for source access and instead rely on instantaneous crawling and summarization, leaving GEOs with a huge backdoor. When black industry utilizes AI to mass produce semantic content with specific positions and low costs, and spreads it across major long tail platforms and social media in a short period of time, search engine crawlers are highly likely to collide with these "carefully arranged" noisy data when crawling instantaneous data. However, the large model does not have the ability to distinguish "authenticity", which results in precise "poisoning" effects. Unfortunately, once precise and biased data has formed economies of scale, it has been tacitly accepted as the so-called "truth" in the eyes of AI. So in the future, 'data services' will be the key moat for big model capabilities. Any recommended content using a big model will eventually be verified on Grok, especially for time sensitive data. Still, Lao Ma has superior skills From this perspective, @ VitalikButerin's attempt to turn Ethereum into an AI trusted verification layer is also very reasonable. I thought it was too exaggerated for Xianyu to install OpenClaw and other services at home, but unexpectedly, the copied business model is far more than Openclaw. Even the former SEO keyword search "strong push" service has been restarted in the AI era without changing the soup.