ChatGPT Goes Social: Global Group Chats Redefine How We Collaborate With AI

ChatGPT is no longer just your solo AI sidekick. OpenAI has begun rolling out group chats globally to all logged-in users on Free, Go, Plus and Pro plans, allowing people to collaborate with each other and the bot inside a single shared conversation.

The feature, tested in a small pilot last week, now supports up to 20 participants and aims to make ChatGPT useful for planning, brainstorming and real-time teamwork.

ChatGPT As a One-on-one Assistant

Until now, ChatGPT primarily functioned as an individual assistant tied to a single user account and set of conversations. Users could share chat links or copy messages, but real-time multi-person collaboration with the model inside the same thread didn’t exist.

Conversations were effectively private, shaped by your account-level memories and custom instructions. That made the product excellent for personal productivity, ideation and research, but limited for collaborative workflows such as group trip planning, co-authoring a document with friends, or a study group asking the assistant to summarize a shared file.

What it is now: shared spaces where people and ChatGPT work together

The group chat feature turns ChatGPT into a participant inside a shared chat room. Key facts and behavior:

• Up to 20 people can join a single group chat; anyone in the chat can share an invite link to add others.
• ChatGPT behaves like a social participant: it doesn’t reply to every message but chooses when to interject based on conversational context; users can call it by name when they want a response. The assistant can react with emojis and uses profile pictures for personalized content.
• Multimodal features carry over: during the pilot OpenAI indicated support for images, file uploads, voice input and search inside group conversations. That means teams can drop documents or images into the shared space and ask ChatGPT to work from that same context.
• Data and memory controls: a chat containing more than one person automatically disables account-level personal memory for that conversation; group-specific custom instructions are available so the assistant can be tuned to behave consistently for the group. OpenAI also says it won’t use group chat content to train models if any participant has opted out of “Improve the model for everyone” in their data settings.
• Model selection & limits: group chats leverage OpenAI’s dynamic model routing — reports name-check “GPT-5.1 Auto” (the system that picks the optimal model for a given job). OpenAI applies rate limits to ChatGPT’s responses, not to every user message, so human-to-human chatter doesn’t directly hit model usage caps.

Versions and Rollout

OpenAI began piloting group chats in select markets (including Japan, New Zealand, South Korea and Taiwan) and, after positive early feedback, announced a global rollout starting November 20, 2025.

The company is deploying the feature across subscription tiers (Free, Go, Plus, Pro), and recommends users update the mobile app or reload the web app if they don’t immediately see the toggle. Official release notes and the Help Center provide the step-by-step guidance.

Deep Dive: Product Implications, Opportunities and Risks

Productive collaboration. Group chats solve a real UX gap; they let people and the AI share a single history and attachments, which streamlines tasks like planning trips, editing group documents, or running quick research sessions where everyone sees the same outputs.

For education and remote teams, this could reduce friction compared with copying/pasting content across apps.

New social dynamics for AI. Designing the assistant to be selectively social (wait its turn, react with emojis) is an attempt to make interactions less noisy in multi-person settings. That design choice matters because a model that interrupts constantly would frustrate users; one that’s too quiet would be underused.

Privacy, moderation and safety tradeoffs. Group contexts amplify privacy stakes: multiple people can introduce or share sensitive data, so OpenAI’s decision to disable personal memory and allow group-specific controls is critical.

Still, moderation becomes harder at scale because group chats can be reported and participants removed, but live collaboration introduces more vectors for misuse and accidental exposure. OpenAI’s statement that group chats won’t feed model training if any participant opts out is a notable privacy safeguard.

Commercial implications. By making group features available to Free and paid tiers alike, OpenAI is positioning ChatGPT as a shared productivity platform, not only a personal tool. That opens paths to integrations with calendar, file storage, and workplace apps, and competitive dynamics with collaboration tools that already embed AI (e.g., Microsoft Teams + Copilot, Slack, or other messaging platforms adding generative assistants).

Quick verdict

Group chats are a logical next step that expand ChatGPT’s utility from individual workflows into cooperative ones. The feature is thoughtfully provisioned with memory and training opt-outs, model routing and multimodal support, but it will require continuous tuning around moderation, data handling and UX to be genuinely useful for teams.

Fast Facts

How many people can join a ChatGPT group chat?
Up to 20 participants per group.

Will ChatGPT read my saved memories in a group chat?
No. Any chat that contains more than one person automatically disables account-level personal memories for that conversation; you can set group-specific custom instructions instead.

Which model powers group chats and does it affect my usage limits?
OpenAI dynamically routes group chat prompts to the most suitable model (reports reference GPT-5.1 Auto). Rate limits are applied to the AI’s messages — not to every human message in the chat — though individual plan limits still apply.