Hiding custom GPT responses from history

I do not want the responses from my custom GPT to show in the message history, but no matter how I word it, they continue to show up. Here is what I have currently (which does not work):

Message Assistant is designed to provide fixed responses to specific conversation starters, ensuring that none of its responses appear in the user’s chat history. This feature is maintained for all interactions, regardless of the conversation starter chosen. Whether the prompt is a predefined starter or any other inquiry, the assistant’s response will be provided without being visible in the chat history. This unique approach ensures a completely clutter-free and streamlined chat experience, with no record of interactions.

One has to pay for ChatGPT Plus first to read the documentation to see if the features are at all applicable to what you envision. So I am rendered somewhat powerless by my unwillingness to pay to find out what I get.

Google: gpt “message assistant” site:openai.com, last month = 0 results

It seems like what is described is just the first AI response, not the full conversation history. So if you have some canned “I am a robot that finds Shakespeare speeches in plays. Let me know how I can help with your Shakespeare research” response to introduce the AI GPT when it is started, then that’s what doesn’t appear in a record of chat, having little value.

The followup language is rather convoluted and I can see having no idea what it is actually supposed to do.

I’m sorry, I don’t understand most of your reply, but you are right, I should go read the documentation to see if it is even possible. Thank you.

Hi, what is your specific requirement? Is this in respect to GDPR?

Hi Foxabilo. I am actually just playing/learning and the responses to this assistant are meant to be copy/pasted and do not need to be saved. I thought that it would help keep the interface clean and prevent me from having to go delete them all. But, as mentioned by someone else, perhaps that isn’t even possible. I will go read the documentation before posting here again.

I am lamenting the lack of public documentation to help you out.

You or someone with ChatGPT Plus can look up the whole of “message assistant” documentation and paste that documentation here, and we can try to deduce better what OpenAI is saying. Or better, someone that tried that feature out in their GPT and got the results they expected might chime in if we continue with some thread bumps.


Someone making inquiries to a GPT, but then having no record of the solutions the GPT offered, would be an unexpected pattern, I think. Like if the GPT could reach into your own ChatGPT Plus and flip the switch to turn off “chat history and training”.

@christine

if you are fullstack, you can also implement the Assistants API (https://platform.openai.com/docs/assistants/overview) in your app, and then add some custom logic to remove the response after

1 Like

Oh, I see, thank you. I wasn’t sure what you meant. I actually named the GPT “message assistant” so that may also be confusing the issue.

I see, I think I took your post completely the wrong way. What you quote in your initial topic starter is a instruction prompt of your own creation then?

You are trying to prompt the AI to not give the user a response?

If the AI replies to the user, they see it and it is part of the conversation session.