From Voice Message to Email Sent
On August 6, 2026, I sent my first email entirely from a voice message.
I simply spoke. A few seconds later, the recipient received a real email with a subject line and the exact message I wanted to send. No keyboard, no manual typing, just my voice.
Just the Gmail plugin in ChatGPT
Technically, this was done using the connection between ChatGPT and the Gmail plugin. I connected Gmail to the official ChatGPT plugin, and that's it.
The first time I tried to send an email using a voice message, ChatGPT asked me to connect my Google Contacts. It failed several times and kept asking me to reconnect them. At one point, I even told ChatGPT that it was repeatedly asking for the same permission. Somehow, even without successfully connecting Google Contacts, it still identified the correct recipient. I dictated the message, it was transcribed into text, I pressed Send, and the email was delivered. It felt surprisingly natural.
Of course, connecting Gmail to OpenAI introduces additional security considerations. I've already made that connection myself, so this is a good reminder not to keep sensitive information or passwords in email. Perhaps it's also another argument for moving towards a passwordless world, where services rely more on one-time passwords and other secure authentication methods.
Now I can send emails while driving
From this point on, I can literally write emails while driving. "Send an email to this person saying this," then another one, and another one. I've already started doing it.
This experience made me think about something bigger.
We are entering a new era where almost any message, email, Slack message, or chat can be generated or assisted by AI.
Welcome to a new era of AI mistakes
In the past, most communication errors came from people. Typos, grammar mistakes, missing words, or awkward phrasing were simply part of human writing.
Now we have a completely new category of errors.
AI can misinterpret a voice message. AI can misunderstand what the person actually meant. AI can confidently rewrite a message to sound perfectly written while subtly changing its meaning.
These are no longer human writing mistakes. They are AI interpretation mistakes.
Ironically, this changes how we perceive imperfect writing. In the past, seeing a typo, inconsistent punctuation, or a message starting with a lowercase letter often looked careless or unprofessional. Today, those little imperfections might become a sign that a real human actually wrote the message.
For the first time, a few small imperfections may signal authenticity rather than poor writing.
I know it's already possible to let AI log into LinkedIn, write posts, reply to comments, and interact with people on your behalf. Personally, I have no interest in that, and I don't support it. I'm responsible for my own reputation, my ideas, and everything I publish. AI can help me capture my thoughts through voice transcription or improve clarity, but it shouldn't replace my thinking or my voice.
I wrote about this a year or two ago, and my position hasn't changed: I don't publish AI-generated posts. Everything I share starts with my own ideas. I do use voice-to-text and transcription, but there's an important difference between helping me express my thoughts and generating them for me.
Talking to AI will soon look completely normal
Voice messages and voice transcription will only become more common, and this experience is another sign of that. For me, the biggest limitation with Claude right now is voice transcription. ChatGPT handles switching between different languages within a single voice message very well. Claude is still less reliable here. So for me, its main imperfection isn't the model itself, but the voice interaction around it. If you speak only English, this is much less of an issue with Claude. And I don't mean an advanced voice interface. I mean the simple interaction of recording a voice message and having it transcribed into text.
It will also be funny to watch people talk to AI more and more. Someone in an office, on the street, or anywhere else will simply start talking into their phone, and you may have no idea whether they're talking to another person or to AI. From the outside, it looks the same.
That's the direction this is all heading: fewer keyboards, more voice, and a growing question mark over who, or what, actually wrote the words in front of you.