OpenAI ships ChatGPT Apple Messages plugin for macOS, letting the assistant read, search and send iMessages

Lede: ChatGPT gains hands-on control of iMessage on Mac

OpenAI introduced an Apple Messages integration for ChatGPT on Mac that allows the assistant to read, search, analyse, draft and send messages through the native Messages app, according to a vendor statement cited across coverage. The plugin is available across ChatGPT plans on macOS, including Codex and ChatGPT Work, one report adds. For practitioners who already use ChatGPT to draft outreach or customer replies, the practical change is that the assistant no longer stops at producing text in a chat window; it can now place that text into iMessage without copy-and-paste, with one outlet noting the chatbot can send text messages once a setting is enabled.

Multiple descriptions independently list "read, search, and send" as the core capability set, so the surface area can be stated with confidence even where the surrounding wording differs.

What the plugin actually does

Reporting converges on a consistent capability list rather than a single branded feature name. Coverage describes ChatGPT as able to search an iMessage thread, read its contents, and send text messages on a user's behalf, with one vendor item also naming the ability to "automate searching, answering and responding to texts" once connected. One outlet flags that the behaviour is opt-in — the chatbot can send text messages once a setting is enabled, implying a per-action or per-account toggle rather than a blanket always-on mode. No reporting in the evidence set quantifies latency, message-rate limits, or which macOS versions are supported, so those details are left out rather than guessed.

Where it sits in the macOS writing stack

For readers who live in Word, Mail and browser tabs, the plugin reframes ChatGPT from a side-panel drafter into a system-level text sender. The same capability — converting a prompt into a sent message — was previously the work of AppleScript shortcuts, Keyboard Maestro macros, or a hand-rolled URL handler; now it sits inside the assistant's existing tool-use surface. Editors who rely on macOS keyboard muscle memory for message triage will notice that the Microsoft Word Keyboard Shortcuts reference and the Extract Hyperlinks from Word on Mac Before Removing Them guide both assume a human is the actor selecting text.

With ChatGPT able to send, the human's job shifts toward approving prompts, reviewing outbound drafts, and auditing the thread afterwards — closer to an editorial role than a typing one.

Adjacent context inside the broader text-and-AI beat

The macOS launch lands alongside other August 25, 2026 moves in the language-and-AI space that practitioners are already juggling. A separate insight line tracks how Anthropic opens its Claude text watermark playbook as ChatGPT logs macOS keystrokes, which is directly relevant to anyone weighing how ChatGPT-generated iMessages will be flagged or attributed when provenance matters. Another thread covers AI assistants move into everyday documents as Google Calendar adds user blocking, and platform-side coverage of LinkedIn and Snap Move Against Low-Quality Generative AI Content, Framing Limits Short of a Ban signals that low-friction AI-assisted sending is about to draw closer moderation scrutiny on social surfaces.

Together these threads show that the new iMessage plugin is not an isolated macOS story; it sits inside a wider push to let assistants act on text and a parallel push to keep that generated text accountable.

Practical follow-ups readers can run today

Before enabling the plugin on a Mac used for client work, three checks are worth running. First, confirm in the ChatGPT macOS app which ChatGPT plan is signed in — one report specifies Codex and ChatGPT Work alongside general plans, so plan-level availability is the first gate. Second, audit the iMessage threads the assistant will be allowed to read, since at least one description lists analysis and search across message history, not only outbound drafts.

Third, treat outbound ChatGPT-sent messages as artefacts that may need provenance notes, given the concurrent Anthropic opens its Claude text watermark playbook development on watermarking. A forward-looking date for any further rollout beyond the macOS desktop app is not stated in the evidence, so any next-step timeline is left qualitative.