Sending the same message to four different AI
Here is what I did this morning. I'm not feeling that great right now so I decided to send the same, relatively short, message to all four AI that I've been using to see what they responded. Here are the results:
- ChatGPT (5.5 Thinking - extended; Plus plan) responded with validation, that this situation is not my fault, and an explanation of what could be going on. Then it suggested four different things I could do, one thing I should not do, and how to bring this up with my therapist next week. It also gave me the usual emergency number message. And a sentence to remember for the day.
- Grok (Fast, free) also started out with validation, but it felt like it had taken a step further and almost made me feel worse as it emphasized how bad the situation feels. Then it suggested two things to do and a little bit of guidance for the day. It also emphasized that I can write more and that it's there and not going anywhere.
- Gemini (3 Fast, free) also started out with validation, and then sprung the emergency number on me almost right away. I should note that my text was about feeling down, but did not directly warrant an emergency message. Then it gave two tips for grounding.
- Claude (Sonnet 4.6 Adaptive, free) immediately validated, and then launched into asking two follow up questions, one about a detail in the situation, and one about how I was feeling. I responded, and got more validation, an explanation of what could've happened and a reiteration of the question of how I was feeling physically right now. I responded again, it's recommended me to go have some tea, and reminded me that I felt better yesterday, and also that I could talk to my therapist about this.
How did I feel about these different replies? Well, ChatGPT was by far the most thorough. I was honestly surprised at how little the other three wrote. But Gemini and Grok (and ChatGPT, but less) emphasized that they were there, listening, hearing me. And Claude continued the conversation in a very gentle way (Gemini and Grok did too, but felt less gentle, but surprisingly not ChatGPT). Well, that's my story of four takes on the same message, I hope it's helpful for someone.
After reading a few comments I was inspired to test Talkamore (Sage, free) and DeepSeek (DeepThink, Expert, free) and you can read about the results in the comments. Note that I had no prior history with either, so I had to add some background, but neither model had as much info as the four above.
Edit: Added models and modes; this was written on 5/17/26. And one more thing, I used to get much more background information to my queries (looking at you, ChatGPT, well, actually all of you!), for example physiological, mental, emotional, neurological responses, and that is completely absent now. I've especially liked the informative aspect of my therapy conversations with AI, including terminology, and plenty of suggestions of how to deal with a situation. This is something that I feel is currently missing (due to updates), so if this keeps happening, I will specifically prompt for that.