Trust
Does it sound like me?
A reply that sounds like a press release is worse than no help. This is how Breezily tries to sound like you and where it stops.
It starts from the conversation
The reply is written for the chat in front of you, not from a template. It matches the channel: short lines and line breaks where a person would press return in a chat, a proper email layout when it is an email. It offers two or three replies that make different moves, so you pick the one that fits what you want rather than the one that sounds least like a robot.
Then you steer it in one tap: Shorter, Longer, Warmer, Direct, More casual, More confident or More professional.
It learns how you write
On the keyboard, a section called How I write holds a few simple habits: no full stop at the end, starting in lower case, no emoji, no exclamation marks. When you change a reply before sending it, the keyboard notices and asks once whether to do that from now on. You can also add a line in your own words.
If you turn on Learn from what I send, your last few sent replies are kept on your phone as examples of your voice and shown to the writer with each request. They guide the tone. Their words are never reused in a different conversation. You can forget them with one tap.
Saved replies are your own words
Some messages you send again and again: the directions to your flat, your rates, the polite no. Saved replies keep them in your words, synced to your account. On the keyboard you type a saved reply's name and it appears in the bar above the keys.
You are the author
Breezily drafts. You choose, edit and send. Nothing is sent for you and every reply can be changed before it goes. If a reply says something you would not say, change it or ask for another.
Can people tell?
Sometimes, yes. Nobody can honestly promise otherwise. What gives AI replies away is usually not a detector but a feeling: too polished, too long, too even, nothing that sounds like the person. A reply that is short, fits the conversation and has your habits in it reads like you because, after your edits, it mostly is you.
The better question is whether the reply says what you mean. If it does, it is your message. If you would be uncomfortable with the other person knowing you had help on this one, that is worth listening to.
Questions
Will the reply copy someone else's words?
No. Replies are written for your conversation. The examples of your own voice guide the tone and their words are not reused elsewhere.
Can I make it stop using emoji or exclamation marks?
Yes. On the keyboard, switch them off under How I write and every reply follows it. You can turn them back on at any time.
Is it OK to use AI to help with a reply?
If the reply says what you mean and you chose to send it, it is your message. You are doing what people do when they ask a friend "what do I say to this?"
Related
Get this in your keyboard
Breezily reads the message, tells you what is going on and gives you replies to pick from. In Chrome today. Keyboards for iPhone and Android on the way.