ChatGPT for iMessage on Mac: What AI-Assisted Texting Means for Builders

OpenAI released a plugin that lets ChatGPT search, draft, and send iMessages directly from a Mac. For AI builders, this is a case study in embedding LLMs into existing communication channels, with all the privacy, consent, and UX challenges that come with it.

ChatGPT lands in iMessage on Mac

The ChatGPT for iMessage on Mac plugin lets ChatGPT search text messages, draft replies, and send messages within the Apple Messages app. It is available in Codex and ChatGPT Work on desktop. OpenAI's head of developer experience, Romain Huet, suggested use cases like identifying frequent contacts, summarizing conversation topics, or surfacing items left unfollowed. The launch follows OpenAI's broader feature push, including a Computer History option that learns from user actions and is off by default.

What AI builders should learn from this integration

For product teams building AI into messaging workflows, this plugin is a concrete example of how to thread an LLM into a private communication channel. The key design decisions: the plugin must be explicitly installed and enabled, and it does not index users' messages. ChatGPT only reads a text when the user prompts it. That opt-in, on-read model is a sensible privacy baseline, but it still requires users to trust that the model won't misuse context.

The public reaction has been sharply divided. Some analysts call it a game changer for productivity, while many users say AI-generated texts cross a line and threaten relationships. Reddit cofounder Alexis Ohanian noted that texting has remained a human channel, and AI integration may accelerate demand for verifiably human communication.

Practical implications for privacy and consent

The plugin's privacy model is straightforward: no background indexing, no persistent access. OpenAI's representative stated that the feature must be explicitly installed and enabled and does not index on users' messages. ChatGPT will only read a text message when the user prompts. Still, users on social media expressed concerns about giving ChatGPT unfettered access to everything. For builders, this highlights the importance of transparent data handling and user control. The Computer History feature, also off by default, shows OpenAI is aware of the sensitivity.

Caveats and unknowns

The current reporting relies heavily on a single major outlet. Details about rollout scope, platform availability beyond Mac, and long-term privacy guarantees may evolve. The feature is currently limited to Codex and ChatGPT Work on desktop; it is not clear if it will expand to other platforms or ChatGPT tiers. The public backlash suggests that even with opt-in design, user trust is fragile. Builders should watch how OpenAI handles feedback and whether Apple, which has ongoing legal tensions with OpenAI, responds.