Companies are pouring over $100 million annually into AI tracking and visibility, but there's a glaring issue: no solid research confirms whether these AI tools provide consistent recommendations for brands and products. Despite the growing reliance on AI like ChatGPT, Claude, and Google AI, the lack of valid metrics raises serious concerns for marketers. This research gap is troubling, especially for executives who might invest heavily in AI tracking without understanding its reliability. To address this, the author and a colleague ran an experiment with 600 volunteers using three popular AI tools, prompting them 12 times and recording nearly 3,000 responses. The results showed extreme variability in the lists generated. For instance, when asking for top chef's knives under $300, the AI provided a wide range of unique brands and products. Statistically, ChatGPT and Google's AI had less than a 1% chance of delivering the same list twice in 100 prompts, while Claude fared slightly better but still struggled with consistency in ordering. The findings suggest that AI-generated lists of recommendations are largely random, which poses a significant issue when consumers rely on them for critical decisions. Despite this randomness, some patterns emerged. Certain brands appeared consistently across multiple runs, indicating that while the lists themselves are unreliable, measuring which brands appear most frequently could provide some insight into their visibility in AI responses. This nuanced understanding could help brands gauge their presence in the AI-driven market, despite the inherent unpredictability of the recommendations.