How to Use AI to Write Better Text Messages in a Second Language

How to Use AI to Write Better Text Messages in a Second Language

I've been texting my Portuguese-speaking in-laws for three years now, and for most of that time, I was basically a robot. Grammatically correct, sure. But stiff. The kind of messages that technically say the right thing but feel like they came from a customer service bot.

Then I started using ChatGPT differently — not to translate, but to help me sound like an actual person. The difference in how my mother-in-law responds now versus six months ago is wild. She sends voice notes. She sends jokes. She stopped using formal pronouns with me. Something shifted.

The Problem With Just Translating

My first instinct was to write what I wanted to say in English, then paste it into a translator. Google Translate, DeepL, whatever. The output was always technically correct.

But here's the thing about text messages: nobody texts formally. In English, I'd text "gonna grab groceries, need anything?" In Portuguese, the translation of that sounds weird. Too complete. Too textbook.

Real Brazilian Portuguese texting involves abbreviations I never learned in Duolingo. Dropped pronouns. Emoji placement that means something different than in English. The rhythm is completely off when you just translate.

I realized I wasn't trying to write correct Portuguese. I was trying to sound like I belonged in a family group chat.

What Actually Works: Context-Loaded Prompts

The breakthrough came when I stopped asking ChatGPT to translate and started asking it to help me text like a specific kind of person.

Here's the prompt structure that changed everything for me:

  • Who I'm texting (relationship, not just language)
  • What I want to say (the actual meaning, not a formal version)
  • The vibe I'm going for (casual, warm, joking, apologetic)
  • Any cultural context that matters

So instead of: "Translate 'I can't come to dinner tomorrow' to Portuguese"

I'd write: "I need to text my Brazilian mother-in-law that I can't make dinner tomorrow. I feel bad about it. We have a warm relationship — she calls me filho. I want to sound genuinely apologetic but not overly dramatic. How would a 30-something Brazilian actually text this?"

The difference in output is massive. ChatGPT started giving me options with the little verbal softeners Brazilians actually use. The "aí" and "né" that make you sound human. The emoji suggestions that don't make you look like a boomer.

The Kick: Teaching It Your Mistakes

Here's the part nobody tells you. The real power isn't in the first response — it's in the correction loop.

I started pasting my bad attempts into ChatGPT. Not asking it to write from scratch, but asking it to fix what I already wrote and explain why.

"I wrote this text to my brother-in-law: [my attempt]. It felt awkward when I sent it. What did I get wrong? How would someone actually say this?"

That's when I learned I was using "você" with people who expected "tu." I was ending messages too abruptly — in Brazilian Portuguese, people don't just stop texting, they have little exit phrases. I was using the wrong laughing sound (kkkk, not hahaha).

After maybe 40 of these correction sessions over two months, something clicked. I started catching my own mistakes before I needed to ask. The patterns made sense now. I wasn't learning Portuguese grammar — I was learning how to text in Portuguese. Different skill entirely.

I also discovered that asking "what would make this sound more affectionate without being weird" gave me better results than any translation request. Cultural tone is the hard part. Words are easy.

What Didn't Work

Claude was oddly formal for this task. Everything came out sounding like a polite email. I tried it for about two weeks and my sister-in-law asked if I was okay because my messages seemed "different."

The voice feature in ChatGPT was useless for this purpose. I thought maybe I could speak my message and it would help with pronunciation. Nope. Text messages don't need pronunciation. I needed written conversational Portuguese, which is its own dialect basically.

Also, asking for slang specifically backfires. I once asked for "casual Brazilian slang" and got something that apparently sounded like a teenager from 2015. My wife laughed for a solid minute. Asking for "how someone your age would text" works better than asking for "informal language."

The weirdest failure: ChatGPT occasionally suggested phrases that were technically Brazilian Portuguese but from a different region than my in-laws. Regional differences in texting style are apparently a thing. I had to add "paulistano" to my prompts eventually.

Six months in, I still use it for anything beyond basic messages. But now I use it less to write for me and more to check my instincts. Which feels like actual progress, even if I can't quite explain why some texts sound right now and others don't. Maybe that's what learning a language actually is — developing instincts you can't fully articulate.

My mother-in-law sent me a meme yesterday that I actually understood without translating. I still haven't figured out how to respond to it.

Heads up: Some links in this post may be affiliate links. I only recommend tools I've personally tested. Opinions are entirely my own.

댓글

이 블로그의 인기 게시물

How to Use Claude AI to Organize Your Messy Inbox (Without Losing Your Mind)

I Handed My Entire Summer Trip to AI — Here's the Honest Breakdown After Two Weeks of Testing

The Perplexity AI Feature That Makes Google Feel Outdated