Will confirm when Go plans and Android apps will also get the feature re-added and let you all know here.
Thank you for checking!
1 Like
We're rolling this out for Go now! To confirm, did this ever work on mobile (iOS or Android)? Or is this a feature request?
It always worked on Android (Iām not sure about iOS), until around late October or early November, when it stopped working.
I just remembered: this feature stopped working on both Web and Android app on November 12, with the global release of GPT-5.1. Before that, up until the launch of GPT-5, we could @mentionmention Custom GPTs. @OpenAI_Support
1 Like
Tested this feature today on both the browser desktop version and the updated Android app. With ChatGPT model 5.2, including a custom GPT via @ mention does not apply that GPTās rules or formatting. The included GPT behaves like the base model and is not aware of its own instructions when asked. This worked correctly in model 5.1. There also appears to be a related issue on Android where @ no longer auto-populates GPT options.
Interesting. I havenāt tested this instructions issue on 5.2 yet, since Iām still using 5 and 5.1. To be honest, this has been a major headache since the launch of 5.1, because it seriously disrupts my workflow and delays deadlines with my clients, due to the fact that Iām forced to work on a computer, since this feature has been disabled on Android. Is there any forecast for fixing this on Android? @OpenAI_Support
Still not working @5.1 and lower - both desktop & Android.
Only workaround is reverting to 4.x in an active chat, set the @ mention and switch back to 5.x when done.
I have both plus and professional account(s) and neither of them work.
Also using Dall-E for images and enlarging them by clicking on it doesnāt work: youāll get a black screen.
Found another glitch.
Chart a chat, with an agent or just GPT.
Mention an agent.
Ask for itās name ā it will say itās the initial agent / GPT.
Press ESCAPE, to quit this particular agent.
Ask the initial agent / GPT for itās name.
It will say it the 2nd mentioned agent (so canceling the agent in the chat will activate it).
The system is really full of bugs.
I was able to get this to work (@mentioning custom GPTs) on both iOS and Android (a Google device). If you're still unable to @mention custom GPTs, please confirm the custom GPT is pinned and share specific repro instructions (app version, device, prompt, etc.). Will keep working to repro this.
For the agent name bug, it seems like some GPTs work differently than others (in my example, the Translate GPT didn't answer with its name, but this is likely due to its prompt), but escaping back to ChatGPT did work. Are you able to share a screenshot of your example? Thank you!
2 Likes
Really? How did you manage to mention a Custom GPT inside the chat? I recently changed phones, I used to have a Samsung M54, and now Iām using a Samsung A56, and when I switched I was hoping it would come back, but no luck.
My current ChatGPT version is the latest on the Play Store (1.2025.350) and my Android version is 16 (One UI 8.0), which is the latest. And yes, my GPTs are pinned in the sidebar. I already unpinned and pinned them again to test whether that was the issue months ago, but it wasnāt.
I installed Chrome on Android and I can mention them through the browser, but when the chat gets very long, it lags a lot, unlike the app, which is more optimized. On the app, I still have no solution.
Iāve had this problem for months with no fix, and itās been really problematic for my work. I donāt have any perspective of a solution either. When I opened a ticket through the Help Center, they told me it was a known issue, but they also had no ETA for fixing it.
The @mention system only works on the web UI and is not available on the Android or iOS apps (mentions cannot be used at all on mobile). On the browser, mentions resolve visually and display the Custom GPTās name, but the active model context does not switch. If I start a normal ChatGPT chat or an existing Custom GPT chat and then mention a different Custom GPT, the conversation continues using the originally loaded context instead of switching to the mentioned GPT. As a result, the mentioned Custom GPTās custom instructions, identity, and purpose are not applied, and the model may identify itself as the base ChatGPT. Knowledge files may only appear to load while not actually using their contents. This is the best explanation I can provide, and Iām attaching a screenshot for clarity. This isnāt a major issue for me personally, but I wanted to report it to help improve the experience for everyone.
1 Like
Interesting. Youāre right. I just tested it on the web, which is the only place where I can still use @mention, and GPT doesnāt actually recognize the Custom GPT. It behaves like the default GPT, without applying the instructions set in it.
Another bug found. When is this supposed to be fixed? Weāre already heading into a full quarter without a solutionā¦ @OpenAI_Support