I Let AI Write My Texts for a Week and Nobody Noticed

I should start by saying I felt weird about the whole thing before I even began.

Not morally weird. Not "what has the world come to" weird. More like -- am I cheating at being a person? That kind of weird. The kind where you know you're overthinking it but you keep going anyway.

The experiment was simple. For one full week, every text I sent would start with an AI suggestion. I'd screenshot the conversation, get three reply options, and pick one. Or edit one. Or use the vibe of one and rewrite it in my own words. But the AI had to be involved. Every single time.

Seven days. Every conversation. My crush. My boss. The group chat. My mom. Even my ex, who texted on day four like the universe was testing me personally.

Nobody caught on. Not one person said "that doesn't sound like you" or "are you okay?" or "why are you being weird."

But the reason nobody noticed is not the reason I expected.

Did the AI Actually Write Good Texts?

Sometimes. Honestly -- maybe 60% of the time.

The serious tone was almost always solid. Straightforward, clear, no extra fluff. When my boss asked me to take on an extra shift, the suggested reply was better than what I would have sent, because I would have said "sure no problem" when I actually wanted to push back. The AI gave me a version that was polite but firm. I used it almost word for word.

The funny suggestions were hit or miss. Some landed. Some were the kind of joke your uncle tells at Thanksgiving -- technically structured like humor but missing something essential.

And the flirty ones. Look. The flirty ones were occasionally great and occasionally made me want to throw my phone into a lake. One suggestion to my crush was so smooth I felt embarrassed clicking copy. Another was so corny I stared at it for ten seconds and whispered "absolutely not" to my empty apartment.

The point is -- it wasn't magic. The AI didn't suddenly make me charming. But it did something else.

What Actually Changed?

Here's what I didn't expect. The biggest shift wasn't in what I said. It was in how fast I said it.

I am a world-class text overthinker. The kind of person who drafts a reply, deletes it, screenshots the conversation and sends it to two friends, waits for their analysis, drafts again, and then sends something completely different forty minutes later. I've done this for texts as simple as "what time works for you."

With the AI involved, that loop broke. Not because the suggestions were perfect. But because having three options in front of me gave my brain something to react to instead of spinning in the void.

It's the difference between staring at a blank page and staring at a rough draft. The rough draft is easier. Even if you rewrite the whole thing, you needed it there to start.

By day three, my average reply time dropped from "embarrassingly long" to "actually normal." I was responding to people within minutes instead of hours. And that -- more than any clever wording -- is what made conversations feel different.

Did I Feel Guilty About It?

Day one, yes.

My crush sent a photo of her dog wearing a raincoat and I got the suggestions and picked one and sent it and it was fine and she laughed and then I sat there feeling like a fraud. She thinks that was me. That was kind of me. Was it me? I chose it. But I didn't write it. But I would have said something similar. But would I?

That spiral lasted about an hour.

By day two, I noticed something. I was editing almost every suggestion before sending it. Swapping a word here, changing the punctuation there, adding the kind of dumb aside that only I would add. The AI was giving me a starting point and I was making it mine.

By day five, the guilt was gone. Because I realized I wasn't outsourcing my personality. I was outsourcing the paralysis.

What Happened When My Ex Texted?

Day four. 11:47 PM. "Hey, been thinking about you."

Of course. Of course this happens during the experiment.

If I'm being honest, I would normally have spent the rest of the night crafting the perfect response. Something that communicates "I'm doing well and I'm unbothered but also not cold and I acknowledge the vulnerability of your message but I'm not opening that door again." You know. Simple stuff.

Instead, I screenshotted it. Got three options. One was too warm. One was too harsh. The third was just -- honest. Short. Acknowledged the message without inviting more. I changed two words and sent it.

Took me four minutes. Normally that text would have ruined my sleep.

She replied the next morning with something casual. The conversation was fine. It didn't spiral. It didn't become a thing. And I think that's because I responded quickly and clearly instead of agonizing for hours and sending something overcrafted at 3 AM.

If you've been in that situation -- and I know some of you have been in that situation -- there's a version of what to text back that covers it better than I can here. But the short version is: quick and honest beats perfect and late. Every time.

What About the Group Chat?

The group chat was where the experiment almost fell apart. Because group chats move fast and the vibe is chaotic and half the humor is callback jokes from 2019 that no AI is going to understand.

I learned quickly that AI suggestions work best for conversations with stakes -- your boss, someone you're dating, an ex, a difficult family member. The emotional weight is what creates the overthinking, and the overthinking is what the AI actually helps with.

For the group chat, I mostly used the suggestions as a jumping-off point and rewrote them entirely. Which still counted, technically. The rule was the AI had to be involved. It was involved. It just didn't get the last word.

Would I Do It Again?

Not exactly like this. The "every single text" rule was too rigid. There are texts that don't need help -- "on my way" and "sounds good" and "haha" do not require three AI-generated alternatives.

But for the texts that actually matter? The ones where I catch myself typing and deleting and typing and deleting? Yeah. I'll keep doing that part. Not because I need someone else to tell me what to say. But because I need something to react to when my brain won't produce a first draft on its own.

That's what Vervo actually does, if I'm being precise about it. It doesn't replace your voice. It gives your voice a running start. And for someone who freezes on every reply that carries even a little emotional weight -- which is roughly 40% of my conversations and probably more of yours -- that running start is the whole difference between responding and not responding at all.

The Thing Nobody Tells You About AI and Texting

The thing nobody tells you is that it's not about the words. It was never about the words.

I sent better texts this week. But "better" didn't mean "more clever" or "more impressive." Better meant faster. More honest. Less filtered through seventeen rounds of self-doubt.

The best text you can send is one you actually send. And if you need a rough draft to get there -- from an AI, from a friend, from screaming into the void and seeing what comes out -- that's not cheating.

That's just how some of us are wired. And I'm done feeling weird about it.