ChatGPT Pro Users in the US Can Now Text Their Dot Agents Over iMessage and RCS, and the Agent Texts Back First

OpenAI's always-on agents can now reach users where they already read messages. ChatGPT Pro subscribers in the US can text their dot, the persistent agent OpenAI introduced at DevDay, over iMessage or RCS, and the dot can text them first. "By popular request, you can now text your dot," the ChatGPT account posted on X on October 28.

How it works#

Users can send a dot a task or something to monitor by text, photo or voice note, and it will message them proactively with updates, OpenAI says. Setup is through the dot's profile, under Texting; the dot keeps the same context and plugins it has inside ChatGPT. OpenAI says the feature does not give the dot or OpenAI access to a user's other text conversations.

Digital Trends reported that users verify a phone number and save a contact, that Android users need RCS enabled, and that a dot sends a link to the ChatGPT app when a task needs approval. Users can also tell it to text only for urgent matters.

A third party in the loop#

OpenAI's release notes add a detail the launch post left out: the texts do not travel on OpenAI's own infrastructure alone. OpenAI says it uses Linq, a third-party messaging provider, to send and receive texts for a dot, and that it ran "extensive" penetration testing with its frontier models and outside security firms as part of that partnership. Engadget reported the same arrangement.

That matters more for an agent than for a chatbot. A dot that can act on a user's plugins now takes instructions from a phone number, through a provider the user did not choose, so the security of the channel becomes part of the security of the agent. OpenAI's answer so far is the one-to-one design, the approval link back into ChatGPT, and the testing it describes. It has not published the results of that testing.

Rivals are taking a slower route. Anthropic's Claude Duet, opened a week earlier as a waitlisted research preview, must ask before it sends messages, makes payments, signs documents or deletes anything.

What it does not do#

Texting is one-to-one: no group chats, phone calls or FaceTime, according to Digital Trends. It is a beta limited to ChatGPT Pro in the US. OpenAI has not said when it will reach other plans or countries.

Limits#

• Setup details such as number verification come from Digital Trends and OpenAI's release notes.
• OpenAI has not said what Linq can see or store, or published its penetration-test findings.
• OpenAI gave no message limits, pricing changes or carrier details.
• How proactive messages are triggered, and how often, is not specified.

Sources#

• OpenAI Help Center: ChatGPT release notes, Texting your dot (October 28, 2026)
• Engadget: OpenAI now lets you text your Dot agent (October 28, 2026)
• ChatGPT on X: "By popular request, you can now text your dot" (October 28, 2026)
• ChatGPT on X: setup and privacy reply (October 28, 2026)
• Digital Trends: OpenAI now lets you text your ChatGPT Dot agent to get things done (October 28, 2026)