By Hamza Faisal
OpenAI has begun letting its always-on Dot agent text users directly, rolling out iMessage and RCS messaging as a preview for ChatGPT Pro subscribers in the United States. The move takes the company's personal AI agent out of the ChatGPT app and into the phone's default messaging app, where it can be reached like any other contact and can even text first when it has an update.
Dots, introduced at OpenAI's DevDay event in late September, are persistent agents built on the company's flagship GPT-6 Astra model. They work on long-running tasks in the background, keep context across a user's ChatGPT conversations, and connect to outside apps through plugins. Until now, users could only reach a Dot through ChatGPT itself or workplace tools such as Slack and Microsoft Teams.
How Dot texting works
Setting it up takes a few steps inside ChatGPT: open the Dot's profile, select Texting, and verify a phone number. The Dot then sends a first text, and from there the conversation works like any other thread. iPhone users reach their Dot through iMessage, while Android users use RCS in Google Messages. Plain SMS is not supported.
Users can send their Dot a text, a photo, or a voice note, and ask it to take on a task. The Dot keeps the same memory and plugins it has inside ChatGPT, so anything already connected there, such as email or a calendar, stays within reach. When a task needs approval, the Dot sends a link to review the request inside ChatGPT rather than handling the approval over text. Conversations are one-to-one only: no group chats, phone calls, or FaceTime.
The catch: a middleman handles the messages
The texting channel does not run on a new Apple messaging API, because no such API exists. Instead, OpenAI routes Dot texts through Linq, a third-party messaging provider based in Birmingham, Alabama. That puts a commercial intermediary inside a channel many users will associate with Apple's end-to-end encryption, and the conversations are not end-to-end encrypted the way a normal iMessage chat between two people is.
OpenAI says it has conducted extensive penetration testing of the Linq setup with outside firms, but it has not published the results. The company stresses that only the account owner can text a Dot, and a Dot can only text its owner, and that switching the feature on does not give the Dot or OpenAI access to the user's other conversations. Still, anyone planning to text their agent sensitive instructions should know a relay sits in the middle of the thread.
A crowded race for the always-on agent
The rollout is deliberately narrow. It is a limited, rolling beta for adult, US-based Pro subscribers, and it is not available to Business or Enterprise workspaces. Dot itself never launched in the European Economic Area, Switzerland, or the United Kingdom, so users there cannot try texting either, and OpenAI has given no date for a wider release.
The timing also puts OpenAI head-to-head with Anthropic, whose rival agent, Claude Duet, messages users first with a morning brief and updates from connected apps. Duet is a separate app rather than a text thread, and it remains invite-only, with access opening to US Max subscribers this week and Pro subscribers next. The contrast is sharp: Duet will be included in Claude's $20 Pro plan, while Dot texting requires a Pro subscription. For now, the agent that lives in your inbox is a perk for a small slice of paying users, and its most interesting detail is not the convenience, but the company standing between you and your AI.



