OpenAI just released an Apple Messages plugin that lets ChatGPT compose your text messages.
Not suggest. Not summarize. Actually write and send texts on your behalf to your friends, family, anyone in your contact list.
The fact that this exists is one thing. The fact that people want it reveals something else entirely: we're now outsourcing intimacy to AI and calling it convenience.
What OpenAI Actually Built
The plugin reads your Messages inbox and conversation history. You can ask it to: compose follow-up texts based on previous messages, send those texts directly, delete messages, analyze your conversation patterns, draft responses in your "voice."
It's positioned as a productivity tool. "Don't waste time writing texts," the pitch goes. "Let ChatGPT handle it."
OpenAI claims the feature runs locally on your device. Your message history doesn't get uploaded to their servers (probably). But even if that's technically true, something has fundamentally shifted: you're asking an AI system to represent you in your most intimate communication channels.
Why Users Actually Want This
Writing texts is friction. Composition anxiety is real. Procrastination happens. Responding to difficult messages takes emotional energy.
AI removes that friction instantly. Generate a response in your voice. Send it. Done.
It feels like liberation. One less task taking up mental space. One less moment of paralysis wondering what to say.
But here's what's actually happening: every automation that reduces friction also reduces judgment. When you let AI compose your texts, you're outsourcing a decision about tone, timing, authenticity, and relational accuracy to a system that optimizes for fluency, not connection.
Each time you use it, you're also training yourself to trust the machine more than your own instincts. And that matters.
When We Trust Machines to Be Us
This isn't the same as previous automations.
Automating code means a tool writes functional lines. Email automation means filters and mass replies—still formal, still business-context. Meeting scheduling means an assistant books time on your calendar.
But text messages are different. They're personal. They carry voice. Tone matters. Timing matters. The fact that you took time to compose something matters to the person receiving it.
Letting ChatGPT write that? That means trusting a machine to understand relational context better than you do. It means believing ChatGPT knows what your friend needs to hear. It means outsourcing judgment about authenticity to an algorithm.
And millions of people seem willing to do exactly that.
The question isn't whether it works. It probably works fine. The question is: what are we losing when we stop writing our own messages?
OpenAI's Real Play
On the surface: "We're helping users communicate better."
Beneath that: ecosystem lock-in.
Every message sent through ChatGPT is one more reason to keep using OpenAI. Every successful text composition builds trust in the system. Every reliance on the plugin makes you harder to move away from.
This is identical to Google bundling Search + Gemini + Lens together. It's the same play as Cursor integrating code editor + AI + hosting. Integration becomes moat. The more of your life that flows through one platform, the more dependent you become.
And the data? Every text you let ChatGPT compose is data about how you communicate. Tone preferences. Relationship patterns. What you say to whom. That's valuable.
The Broader Shift
We're in the middle of a quiet transformation: human communication is becoming AI-mediated.
Slack helps you draft work messages. Email has AI helpers. Texts now have ChatGPT. Social media suggests your responses. Even calls might eventually have AI assistance.
The trajectory is clear: fewer direct human-to-human interactions. More: AI sitting in the middle, mediating, translating, composing.
What happens when all your communication is filtered through systems designed to optimize for engagement and retention? What happens to authenticity when every message is algorithmically smoothed?
What This Reveals
That people will trade authenticity for convenience. That we're willing to outsource intimacy if it reduces friction enough. That we trust machines to represent us in moments that matter.
And that platforms understand this perfectly. They're betting that once you start using AI to write your texts, you'll use it for everything else too. They're building lock-in not through features, but through trust.
This is the deepest incentive misalignment yet: users seeking convenience, platforms seeking control, and both paying the cost of what gets lost when AI mediates our most human connections.
For more on building, visit Bitroot.