Grok has X, Musk, and massive compute. It still has not become essential. That is the real issue. Not ideology. Not branding. Not even raw model quality, at least not first. Grok’s downloads fell to about 8.3 million in April from more than 20 million in January, while paid usage stayed almost flat at 0.174% versus 0.173% a year ago. ChatGPT, by comparison, had more than 6% of surveyed respondents saying they paid for it. That gap says a lot. Distribution was supposed to be the cheat code. Grok sits inside X, one of the most valuable real-time information networks in the world. It has Elon Musk’s brand behind it. It has the kind of infrastructure story most AI startups can only pitch in fundraises. And yet the product still has not created a clear “why this over ChatGPT, Claude, or Gemini?” answer. That is a brutal product-market signal. The best line in the piece was the engineer who said: “OpenAI is Coke, Anthropic is Pepsi and Grok is RC Cola.” Crude, but directionally right. Grok is visible. It is not indispensable. That matters for the AI trade. The market keeps treating AI as a scale race, but the consumer layer is becoming a habit race. The winners are not just the models that exist everywhere. They are the models users reach for without thinking. ChatGPT has that. Claude is building it in coding and knowledge work. Gemini has Google distribution and enterprise pull. Grok still feels more like an X feature than a standalone workflow. The second-order problem is that Musk can win the infrastructure side while Grok loses the product side. SpaceX renting major data-center capacity to Anthropic tells you compute itself is valuable, maybe more immediately monetizable than chasing chatbot share. But it also sharpens the question: if Grok is the priority, why is a rival getting key capacity? Bottom line: Grok’s problem is not that nobody has heard of it. The problem is that too few users have decided they cannot live without it. In AI, distribution gets you trial. Relevance gets you retention.