Group ChatGPT

Launching a new chat app is just about the hardest thing you can do in tech. And with each passing year, as the incumbents are more entrenched with more users, it gets harder. In 2025, everyone already has their chat app of choice and it's not just getting one user to switch, it's getting that user and all their friends/connections they wish to chat with to switch. Even if they're able to pull over a few people, it will fragment the connections, now requiring people to use multiple apps. These days, many of us may use two or three chat apps (or more!) for different purposes (work, friends, etc) but even those are pretty well established at this point.

Of course, it's slightly easier if you're not doing it from scratch. Which is why apps like Spotify and now Xitter keep trying to do it. The thought is if hundreds of millions of people are already using your app, why not let them do the most social thing of all (at least online): chat? This has worked to some extent with Instagram, but that was also in part because Meta already had other chat apps and could leverage those graphs/connections. But for the most part, layering in chat to an app doesn't really work because it's not really the core function of the app and it feels tacked-on at best. Now OpenAI is going to try their hand at this:

Today, we’re beginning to pilot a new experience in a few regions that makes it easy for people to collaborate with each other—and with ChatGPT—in the same conversation. With group chats, you can bring friends, family, or coworkers into a shared space to plan, make decisions, or work through ideas together.

Whether you’re organizing a group dinner or drafting an outline with coworkers, ChatGPT can help. Group chats are separate from your private conversations, and your personal ChatGPT memory is never shared with anyone in the chat.

This is a pretty limited test, only available in Japan, New Zealand, South Korea and Taiwan.¹ But I suspect it will expand fairly quickly, because I suspect it might actually work – at least to some extent. The reason why here should also be obvious: ChatGPT is already a chat app.

Sure, it's not a chat app in the traditional sense in that to date you haven't been chatting with people, but you have been chatting with AI. And it's by far the most popular chat-with-AI service. It's a chat app quickly approaching a billion users.

In that sense, it's almost strange that you can only chat with AI and not the other people you know that are clearly using it. So that's what OpenAI is now enabling.

Granted, this being-social-with-humans element is not the normal behavior of the service to date, so the already-large user base could reject it. But just as likely is that it makes ChatGPT more sticky because you can now not only chat with AI, but with your friends too, and rope in that AI as needed/wanted.

The main problem may be getting the UI/UX correct for this shift. You can see a world in which trying to make ChatGPT more traditionally social muddles the message, quite literally. And there's already an increasing amount of stuff going on in ChatGPT as OpenAI keeps adding in more functionality. But if they can keep it relatively simple, I could see this working.

And actually, group chatting is the exact functionality that I felt was missing from OpenAI's other app, Sora, from the get-go. The TikTok-like social feed has been fun, but once the novelty wears off, it's going to be more fleeting for many people. The real key to Sora are the "cameos" – inserting people into the videos. This has worked at viral scale for famous people, but that's obviously problematic for OpenAI as well. "Regular" people are far less problematic (assuming they opt-in to letting cameos be created of course!) but it's always going to be weird for many people to post those to an open-ended feed that anyone can follow. Having the ability to make and send these types of Soras to smaller groups makes a lot of sense.

And clearly OpenAI knows this as it's one of the first major tweaks they've made to the app. Rather than default to posting to everyone, the share flow now includes picking people/groups to send your Soras to. That's good, and it could be even better as a feature of, say, a social layer of ChatGPT itself with it's nearly one billion users.

Of course, the biggest push back on all of this will be the privacy/safety angle. (OpenAI's help page on the matter is actually quite informative and straightforward – notably, they can use the contents of the chat for training unless one person in the group has opted out.) And that's why OpenAI is leaning heavily on the messaging around that aspect, as they reiterate in their post:

Group chats are separate from your private conversations. Your personal ChatGPT memory is not used in group chats, and ChatGPT does not create new memories from these conversations. We’re exploring offering more granular controls in the future so you can choose if and how ChatGPT uses memory with group chats.

You’re in control. You have to accept an invitation to join a group chat. Everyone can see who’s in the chat or leave at any time. Group members can remove other participants with the exception of the group creator, who can only be removed by leaving themselves.

Additional safeguards for younger users. If someone under 18 uses group chats, ChatGPT automatically reduces exposure to sensitive content for everyone in the chat. Parents or guardians can turn group chats off entirely through parental controls⁠(opens in a new window).

Of course, users have long been fine entrusting Meta with such data so the question is if the AI angle in particular changes that equation in some way. But obviously Meta has also already inserted AI into their own chat apps from WhatsApp on down too, it's just not the main/leading functionality of those apps.

In that light, this is interesting/risky:

We’ve also taught ChatGPT new social behaviors for group chats. It follows the flow of the conversation and decides when to respond and when to stay quiet based on the context of the group conversation. You can always mention “ChatGPT” in a message when you want it to respond. We’ve also given ChatGPT the ability to react to messages with emojis, and reference profile photos—so it can, for example, use group members’ photos when asked to create fun personalized images within that group conversation.

One imagines that some people won't like the idea of AI just "lurking" in their chats. But you can also see a world in which that's actually fun and potentially useful. This will be pretty polarizing, I imagine!

I also imagine that OpenAI would love continuing to show up Meta on their own turf: social. Just as Mark Zuckerberg is betting his entire company on trying to compete in AI.

One more thing: while I don't hit on it above, there are clearly some potentially interesting business use cases with ChatGPT groups as well. Ones that could be far more lucrative for the company, which seems important going forward...

¹ A VPN is your friend if you wish to try it out – but yes, your connections, unless they're in one of those countries, will have to VPN-in as well. ↩