OpenAI's Dot Agent Can Now Be Texted Through iMessage and RCS
OpenAI has begun rolling out a preview that lets US Pro subscribers text their Dot agent through iMessage on iPhone and RCS on Android. Users can send texts, photos and voice memos to a Dot through the feature, and a Dot can contact its user first with text updates about tasks assigned to it.
Dot was introduced on September 29, 2026, alongside a new $500 paid tier. OpenAI CEO Sam Altman presented the agent as remarkably capable and always-on. OpenAI describes Dots as proactive assistants that keep working across complex projects and everyday tasks, and as always-on agents that can research, analyze data and prepare documents. Bloomberg reported in October that Dots is designed for OpenAI users who already rely on its products at work for writing code and producing research.
Setup happens inside ChatGPT. A user opens the Dot's profile and selects Texting. The feature needs iMessage enabled on an iPhone, or RCS (Rich Communication Services, the upgraded texting standard that replaced basic SMS on many Android phones) on an Android device. OpenAI partnered with Linq to provide the messaging support behind it.
Text is one of several channels. OpenAI states that users can message or call their Dot within ChatGPT, and can type messages while talking to it. Bloomberg also reported at launch that a Dot can be contacted by sending a message on Slack.
Memory ties the channels together. Information a Dot writes to its memory is shared across ChatGPT and text, so context carries over between conversations. A user who sets up a task in the app and follows up by text is therefore talking to the same agent with the same stored context.
OpenAI's separate Messages integration works differently. Its ChatGPT release notes describe an Apple Messages plugin in the ChatGPT desktop app that can read and search iMessage, SMS and RCS conversations and prepare or send messages through Messages. The plugin documentation says that in Codex and ChatGPT Work, the integration can read and search those chats on a Mac and send messages on behalf of users. That integration runs on the Mac against the user's own message history, which is a different arrangement from a Dot holding its own thread with the user.
In my view, the notable change is the interface. An agent that lives in the message thread a person already uses asks for no new app, no dashboard and no new habit. Delegation then starts to look like sending a message to a colleague, which is a lower bar than opening a console. The questions I would check first sit in the details: how an agent-initiated text is distinguished from a user-initiated one, what an agent is permitted to send on a user's behalf, and how the memory store is scoped and surfaced when a conversation spans several channels.
The broader context here is familiar to anyone who has tracked messaging platforms. Each time a service has moved into SMS-style threads, from early chat clients to mobile push notifications, the convenience gain has come with a new surface for notifications, consent and mistakes. Dots add a further wrinkle, because the agent can initiate contact on its own about work it has been given. If the feature works as described, a long-running agent could feel like an ordinary part of a person's day rather than a tool they have to remember to open. That is the part I would watch, along with how the controls around it develop.
