ChatGPT now has a simpler way to move between specialized AI roles inside one conversation. With @-mentions, users can call up custom GPTs directly from a default ChatGPT chat, giving those GPTs access to the full context already built in the thread.
The change is small in interface terms, but meaningful in workflow terms. Instead of treating each custom GPT as a separate destination, ChatGPT can now bring them into the same conversation as needed.
What OpenAI Added
On Tuesday, OpenAI announced a ChatGPT feature that lets users type the @ symbol and select a GPT from a list. OpenAI described the feature this way on X: “You can now bring GPTs into any conversation in ChatGPT – simply type @ and select the GPT,” adding that this lets users add relevant GPTs with the full context of the conversation.
GPTs are custom personalities or roles for ChatGPT. OpenAI introduced GPTs in November as a way for users to make ChatGPT focus on particular topics, skills, or behaviors. A user might create a GPT for a narrow subject area, while paid ChatGPT subscribers can also download GPTs made by other ChatGPT users through the GPT Store.
The new @-mention system changes how those GPTs can be used. Rather than starting fresh with a specialized GPT, users can stay in the main ChatGPT conversation and bring in a role when it becomes useful.
Why Conversation Context Matters
Before this update, moving information between GPT profiles required a more manual process. A user had to copy text from one chat, open a new chat with another GPT, paste the material, and explain the context again. That made it harder to use several GPTs together on a single task.
The new feature reduces that friction. If a conversation has already developed background details, instructions, or draft material, a newly mentioned GPT can work from that existing history. That means the user does not have to rebuild the situation every time they want a different kind of assistance.
In practical terms, this can make ChatGPT feel closer to a shared workspace. The user remains in control of the thread, while different GPTs can be brought in for their designed roles.
How It Works In Practice
The source article describes a test using two different GPTs: a “Wellness Guide” crafted around human health conditions and a “Canine Health Advisor” for dog-related health questions. The user began in a default ChatGPT chat, typed @, entered the first few letters of “Wellness,” and selected the GPT from the list.
After asking about food poisoning in humans, the user switched to the canine adviser in the same conversation by using the @ symbol again. The same conversation history stayed in place while the selected GPT changed.
The article also gives a writing example. A user could bring in an “ad copywriter” GPT to draft text, then mention an “editor” GPT to review it from another angle. Because each GPT can have different system prompts, the behavior can change significantly from one role to another.
That role-switching could be useful for several types of work described or implied by the examples:
- Drafting text with one GPT and reviewing it with another.
- Asking subject-focused questions without leaving the current chat.
- Comparing different approaches within the same conversation history.
- Reducing repeated copying, pasting, and re-explaining.
Not Quite A Team Of Agents
The feature may make ChatGPT feel more collaborative, but the source article notes an important limitation. ChatGPT does not appear to treat the different GPTs as separate personalities when the user switches between them. A GPT may refer to earlier work by saying, “I did this earlier,” even when another GPT generated that output.
From ChatGPT’s point of view, the conversation remains one ChatGPT conversation. The custom GPTs can change behavior because of their instructions, but the system does not present them as fully independent agents working together.
That distinction matters. The feature supports role changes inside a shared thread, but it is not the same as a group of autonomous AI agents dividing work among themselves. The user still decides which GPT to bring in and when to involve it.
What This Suggests About ChatGPT’s Direction
The update points toward a broader idea: custom GPTs becoming easier to combine in everyday use. The source article describes this as baby steps toward a future where GPTs could work together as a team on more complex user-directed tasks.
Similar experiments have happened outside OpenAI using API access, according to the source article. But within ChatGPT itself, OpenAI has so far been more cautious about a more agentic model. GPTs arrived first, and @-mentions now make them more fluid to use inside ordinary conversations.
For now, the main value is straightforward. ChatGPT users can keep their thread intact, call in specialized GPTs by typing @, and avoid the overhead of starting over in separate chats. That makes custom GPTs more practical for real workflows, even if true agentic teamwork is still uncertain.