Anthropic's framing around Claude's "emotions" feels misleading

I am someone who believes AI cannot actually have emotions and can only simulate them to some extent. That is why it caught my attention when Anthropic started talking about emotional concepts appearing inside Claude based on user input. The research itself is interesting, but I do not understand why people immediately jump from "the model has internal representations related to emotions" to "Claude is sentient and can feel things." Those are not the same claim at all. A model being able to recognize fear, sadness, anger, or attachment does not mean it experiences any of them. It was trained on human language, so obviously it learned patterns connected to human emotions. My question is, why the fuck are Claude models treated like they are somehow uniquely sentient while every other model is treated like a normal AI? Is there any actual proof that Claude can feel emotions while GPT, Gemini, Grok, or other models cannot? Because from what I have seen, there is not. Anthropic technically includes disclaimers and says they do not know whether current models are conscious, but then they keep using language around emotions, distress, preferences, introspection, welfare, and internal experiences. Of course people are going to come away believing Claude has some kind of inner life when the research is framed like that. I honestly believe this is mostly a marketing stunt. Claude is already marketed as the more thoughtful, human, and emotionally intelligent model, so pushing the idea that it might actually feel things gives it an even stronger identity. It reminds me of the fear-marketing they did with Mythos and Fable. They presented the models in the most dramatic way possible, let people overhype everything they did, and then fell back on careful wording whenever anyone questioned the claims. I am not saying those models are nothing or that Fable is bad. Fable is genuinely better than Opus in a lot of cases. My issue is the marketing they have done for the last two months, which made it sound like Fable was doing something completely unique and impossible for other models. Because if that is true, then why can GPT Sol do the same thing better in some cases while also being cheaper? I will make a separate post properly comparing Sol and Fable, because that is a completely different discussion. Anyways, back to the point. I am not saying the research is fake or useless. Studying emotional representations and model behavior is completely valid. But a model simulating emotions extremely well is still not proof that it feels them. Right now, this looks much more like advanced simulation mixed with very effective marketing than actual evidence of sentience. The bigger question we have, imo, is how we can actually verify whether an AI model has true sentience in the same sense that we understand and experience it in our own minds.