AI Products
ChatGPT’s Mac app adds an Apple Messages plugin that can search conversations and draft replies
OpenAI has reportedly rolled out an Apple Messages plugin for the ChatGPT desktop app on macOS, letting users search conversations, prepare replies and send messages with confirmation.
OpenAI’s ChatGPT desktop app for macOS has reportedly gained a new Apple Messages plugin, turning the assistant into a more direct participant in everyday communication on the Mac. 9to5Mac and TechCrunch reported on August 20 that the plugin lets ChatGPT search message history, read relevant conversations, draft replies and send messages through Apple’s Messages app after user confirmation. The rollout appears in the ChatGPT Mac app rather than inside Apple’s own Siri or Messages interface, which makes it a notable example of an AI assistant extending into a sensitive operating-system surface.
The practical appeal is straightforward. People often use chatbots to compose difficult replies, summarize a conversation or find context buried in a long thread. A Messages plugin reduces the copy-and-paste work that usually sits between the assistant and the messaging app. A user could ask ChatGPT to find a recent plan, prepare a response in a specific tone or compare earlier messages before writing back. Reports say the plugin can work with iMessage, SMS and RCS messages handled by Apple Messages, with the assistant operating from the Mac desktop app.
The confirmation step is important because messaging is a high-trust action. Sending the wrong message, choosing the wrong recipient or exposing private context would be far more serious than generating a draft in a blank chat window. TechCrunch reported that the feature requires the user to approve the message and recipient before sending. 9to5Mac also noted that users can decide whether to grant one-time permission or persistent access, a distinction that will matter for privacy-conscious users and for anyone sharing a computer or handling work communications.
OpenAI’s broader direction is visible in the design. ChatGPT is no longer being positioned only as a place to ask questions. It is becoming an agentic desktop layer that can act across apps, use local context and turn conversation into workflow. Recent desktop features have moved in the same direction, including support for work apps and coding environments. Messages is different because it touches personal relationships as well as productivity. That makes the user experience less about raw model capability and more about consent, transparency and reversible control.
The timing also highlights a larger platform question for Apple. Apple has been working to make Siri and Apple Intelligence more capable, but outside assistants are moving quickly through app integrations on the Mac. A third-party assistant that can read and act on Messages may become useful before Apple’s own assistant fully coordinates similar tasks. At the same time, Apple’s permission model gives users a familiar gatekeeping layer, which may help outside AI tools work with personal data without bypassing operating-system controls.
There are still limits and risks. Reports indicate that the plugin is available on Apple silicon Macs, leaving Intel Mac users out for now. The feature also depends on user trust in both the local permission prompts and OpenAI’s handling of message context. Even when processing is initiated by the user, message threads can contain addresses, health details, family issues, business negotiations and other information that people would not casually paste into a chatbot. Enterprises will need to decide whether such plugins fit their security policies.
The significance is not that ChatGPT can write a text message; many people already ask AI for help with wording. The shift is that the assistant is moving closer to the point of action. If users become comfortable approving drafts, searching chats and delegating routine communication tasks, desktop AI products could become a layer between intent and apps. If the permission experience feels too broad or unpredictable, users may keep using AI as a drafting tool while refusing to connect it to the inboxes and conversations that matter most.