After today’s ChatGPT interface update on June 24, I noticed a change in the behavior of the “@” command.
The updated interface now shows a more polished selection menu with icons for tools such as plugins, actions, MCPs/connectors, and related options. However, before this update, typing “@” also displayed my available custom GPTs in the same selection list, allowing me to quickly invoke a custom GPT inside the current conversation.
Now, custom GPTs no longer appear in the “@” selection menu.
Expected behavior:
When typing “@”, the selection menu should continue to include the user’s available custom GPTs, as it did before the update.
Current behavior:
The “@” menu shows tools, plugins, actions, MCPs/connectors, and similar options, but it no longer lists custom GPTs.
Impact:
This makes it harder to use custom GPTs inside an existing conversation. Users now need to manually open the GPT from the GPT list or start a new conversation with it, instead of invoking it directly with “@”.
Could you please confirm whether this change was intentional, or if this is a regression introduced in the June 24 interface update?
Here is another screenshot showing my custom GPT, “UI/UX Designer,” being used after I invoked it in the chat through the previous prompt by using the “@” command and selecting it from the GUI.
Although the invocation itself does not appear in the conversation history, the interface clearly returned and applied the selected custom GPT option, confirming that this behavior was previously supported.
After today’s ChatGPT interface update, my Custom GPTs no longer appear when I try to invoke them inside an existing conversation using @.
This is not just a minor UI inconvenience. It breaks an important workflow.
The reason I use Custom GPTs inside an existing conversation is because they can receive the current conversation context. Opening a GPT separately from “Explore GPTs” or “My GPTs” starts a new chat, which does not preserve the context of the conversation I was already working in.
For my workflow, this changes everything.
Before:
I could work in one conversation.
When needed, I could call a specialized Custom GPT with @.
The GPT had access to the current context.
I could continue the work without manually summarizing, copying, and pasting context.
Now:
My Custom GPTs no longer appear through @.
Opening them separately creates a new conversation.
The current context is lost.
I have to manually create context handoffs, which is slower, error-prone, and not equivalent.
This feels like a regression, not just a redesign.
Can OpenAI clarify whether this is intentional, a temporary rollout issue, or a bug?
If this change was intentional, please consider restoring a way to invoke Custom GPTs inside the current conversation with full conversation context. For users who rely on multiple specialized GPTs as part of a workflow, removing this capability is a major step backward.
After the ChatGPT interface update on June 24, I am no longer able to call GPTs using @mentions during a chat.
Before this update, I could type “@” in a conversation and select one of my GPTs to use within the same chat. After the update, this functionality appears to be broken or unavailable.
In addition, while I was chatting inside a Project, the entire Project suddenly disappeared. This is a serious issue because it affects ongoing work and makes the interface feel unreliable.
Issues observed:
GPTs can no longer be invoked via @mentions during a chat.
A Project suddenly disappeared while I was actively chatting inside it.
These problems started after the June 24 ChatGPT interface update.
Given the severity of these issues, I believe this update should be rolled back immediately or at least partially reverted until these bugs are fixed.
Same here - I have four clients that use GPT workflows and none of them work today. This was the selling point to make them purchase GPT Business! Without it, everything breaks. If this was a product decision, it was a bad one. Please fix asap!
This is the whole reason many GPT Business accounts subscribed - to create workflows where context can be shared from one GPT to the next. Business has built corporate workflows off of this - you cannot take away such important existing functionality that was marketed without informing your user base far in advance!
I’m having the same issue with my custom GPTs, and it has wrecked several workflows. What is the status of this and will it continue this way for the immediate future? Want to know if I need to find a workaround or wait until a fix is issued.
