ChatGPT can now read and send your Apple Messages. What it changes for your business

Written by

in

Two chat bubbles on a dark green background, the lower one amber, linked by a dotted line to a glowing mint spark icon.

ARTICLE

ChatGPT can now read and send your Apple Messages.
What it changes for your business

On 20 August 2026 OpenAI added an Apple Messages plugin to the ChatGPT desktop app for Mac. It can read, search and send your iMessage, SMS and RCS conversations. It is available on every plan, and it works in only two specific places, neither of which is a normal ChatGPT chat.

On 20 August 2026, OpenAI announced an Apple Messages plugin for the ChatGPT desktop app. In plain terms: on an Apple silicon Mac, ChatGPT can search your text conversations, write a reply and send it through the Messages app, after you approve the message and who it is going to.

That is a real change for anyone who runs a business over text. Plenty of UK trades, salons and one-person consultancies do most of their customer admin in Messages, and none of it has ever been searchable by anything other than a thumb. The part most of the coverage skipped is where the plugin lives. It does not work on your phone, it does not work on the web, and it does not work in an ordinary ChatGPT conversation.

What actually changed?

OpenAI released a plugin, which is an add-on you install into ChatGPT to give it access to something else on your machine. This one connects it to Apple’s Messages app. In OpenAI’s own words, in Codex and ChatGPT Work it “can read and search iMessage, SMS, and RCS chats on your Mac and send messages on your behalf through the Messages app”.

So you can ask it to find the thread where a customer agreed a price, summarise a week of back and forth with a supplier, or draft the reply you have been avoiding since Tuesday. Both the reading and the sending happen on your Mac.

Source: OpenAI, ChatGPT release notes, 20 August 2026, and the Apple Messages plugin documentation.

Where does it actually work?

In two places only: Codex and ChatGPT Work, both inside the ChatGPT desktop app for macOS. OpenAI is unusually blunt about the rest. Its documentation says the plugin “doesn’t let you interact with ChatGPT remotely through Messages, and it doesn’t work in regular ChatGPT chats”.

There is a hardware limit on top of that. For this release the plugin ships only in the Apple silicon build of the desktop app, so a Mac with an Intel chip cannot run it. It is also absent from ChatGPT on the web and on mobile, the Codex command line tool and the code editor extension.

If ChatGPT Work is unfamiliar, it is the mode that lets ChatGPT act across the apps on your computer rather than just talking to you in a box. We covered what it does in our explainer on ChatGPT Work for small businesses, and Google’s equivalent in what Gemini computer use means for your business.

Does this cost extra?

No. OpenAI’s documentation states the Apple Messages plugin “is available on all plans in the ChatGPT desktop app for macOS”, which includes the free tier. That is consistent with the direction of travel this month, after OpenAI widened what free accounts can do in early August. The cost here is not money, it is setup: an Apple silicon Mac, the desktop app, the plugin, and the macOS permissions granted before it can read anything.

Can it send a text without asking me?

By default, no. ChatGPT asks you to approve both the message and the recipients before anything leaves your Mac, and you can choose “Allow once” to approve only that one send.

There is a second option, and it is the one to think about. If you select “Always allow sending to this chat”, ChatGPT can send future messages to that conversation without asking again. OpenAI’s own warning is worth reading twice: “Persistent approval removes your final chance to review a message before ChatGPT sends it as you.” You can undo it later in Settings, under Computer use, by removing the chat from the “Always allowed to send” list.

Is it safe to let Ai read customer messages?

It depends entirely on which chats you point it at. OpenAI advises keeping per-send approval “for chats that may contain untrusted or misleading instructions”. That is a plain warning about a real problem: a message from a stranger can contain text written to trick the Ai into doing something you did not ask for, and if that thread is on permanent approval, the reply goes out under your name.

The reassuring half is that this runs on your Mac rather than in a data centre somewhere, and if your business uses a managed ChatGPT workspace, an administrator can switch it off through the existing Computer Use control. One fault to know about: if a task is set to full access, or otherwise has approval prompts turned off, the send confirmation may fail to appear. OpenAI’s fix is to switch back to an approval setting and try again.

What this means for your business

Picture a two-person plumbing firm that books almost everything by text. On a Friday afternoon they open ChatGPT Work on the office Mac and ask it to list every customer message from the week that never got a reply, then draft an answer to each one. They read the drafts, correct two, approve them one at a time, and clear an inbox that normally eats a Saturday morning. That is the shape of the win: not clever, just the admin nobody has time for.

The cost is nothing on top of what you already pay, so the real question is effort. Budget half an hour for setup, and be honest about the hardware: no Apple silicon Mac, no plugin. The caveat is the approval setting. Leave it on per-send approval. The moment you let it send to a chat without asking, you have handed your customer relationships to something that cannot tell a genuine request from a manipulative one.

What to ignore

Ignore the headlines that read “ChatGPT can now send texts for you” over a picture of a phone. This is a Mac desktop feature. It will not appear on your iPhone, and typing a request into a normal ChatGPT chat will get you nowhere.

Ignore, too, the idea that you can now text ChatGPT and get answers back the way you would message a colleague. OpenAI says outright that the plugin does not let you interact with ChatGPT remotely through Messages.

What should a small business do next?

If you have an Apple silicon Mac, this is a twenty-minute experiment rather than a project. Install the ChatGPT desktop app, open Plugins, install Apple Messages, then start a ChatGPT Work chat and ask one small, safe question: find every message from a named supplier in the last month and summarise what was agreed. That tells you within minutes whether it is accurate enough for your business, before you let it write anything.

If you do go further and let it draft replies, keep per-send approval on and start with a handful of familiar customers rather than everyone in your address book. On an Intel Mac or a Windows machine there is nothing to do this month, and the capability will still be there if it widens.

Where Breezi fits

A new plugin is a good afternoon. A plugin that quietly clears the admin you have been putting off since March is the point. Membership gives you bespoke hours with a senior Ai team every month: you bring the problem, we build the prompt, workflow or automation that handles it. Lahat Creative, a UK marketing agency, reclaimed up to 30% of their working hours every month doing exactly that. Consultancy does the same at the scale of a team or a whole organisation, the way we did for the financial platform Uphold.

Free guide

Not sure where to start?
Get the cheat sheet.

ChatGPT, Claude, Gemini, Microsoft Copilot, Perplexity, and NotebookLM. What each one is for, how to set it up, and what to do in your first hour. Plain English. Ten pages. No fluff.

    You’ll also get the occasional email about Ai that’s actually worth reading. Unsubscribe whenever.