This is my fourth year using large language models a lot. Lately, I've noticed something: I lose patience with AI-generated text very quickly. I skim through it fast, and I'm often not satisfied with what it gives me.
At first, I thought this impatience was just because AI content feels repetitive, or too safe and balanced. But recently, while using AI to polish my English, my second language, I realized something different. What bothers me is not that AI writes badly. It's that sometimes it writes too well.
By "well", I mean fluent, complete, and polite — like someone who always knows the right way to communicate. But that's not how people really talk in real relationships.
Real people hesitate.
They repeat themselves.
They make grammar mistakes.
Their emotions matter more than their logic sometimes.
Here is what happened.
One day, I needed to explain a difficult situation to someone who speaks English natively. I had so much to say, and it felt hard to organize, so I talked to AI first and asked it to turn my thoughts into four or five clear sentences. When I read the result, it felt complete. So I sent it. But something I didn't expect happened: the other person understood the content, but felt very disappointed — because my words sounded like I was cold, rational, withdrawing from them.
I was confused. My intention with this self-disclosure was to help the other person understand me better, so they could make the right decision — not to make them think I was pulling away, or to cause them emotional pain. So while I was reading their message, I started crying. I am a very emotional person — my feelings show easily on my face. When they tried to call me, I said I was crying and couldn't talk right now. I explained that I wasn't losing control, I was just expressing my emotions in my own way. I also told them I never meant to make them sad.
Looking back, that message I typed in the moment wasn't "polished". Some parts weren't natural, and there were grammar mistakes. But it sounded like me. It kept my confusion, the traces of my second language, my emotions in the moment — and the clumsy effort of me trying to explain myself to someone else. In the end, the other person understood what I meant, and the conversation continued.
So when I looked again at those four or five AI-generated sentences, they were complete, clear, and polite. They described my situation well, balanced the cultural differences, and even included an apology and some understanding of the relationship. But reading them again, I felt strong impatience. Because that message sounded like a relationship manual — not like something a real person, confused and overwhelmed under pressure, would actually say.
That's when I understood what really bothers me: AI's optimization might be quietly taking over what I call my "right to express myself". It doesn't just make my English smoother. It might also be deciding, without me realizing it, what kind of emotional expression counts as "mature", what kind of vulnerability counts as "appropriate" — and what version of me is allowed to be seen.
And this is exactly what social work cares about. Social work values human dignity, self-determination, diversity, and connection.
What social work notices
Dignity means my imperfect language still has value. It shouldn't be reduced to one "correct" answer.
Self-determination means I have the right to decide how I express my emotions, not AI.
Diversity means the imperfections of a second language, the hesitations from cultural differences, and different ways of expressing emotion are all part of being human.
Connection means that words in a relationship aren't only there to sound mature or appropriate. They exist so the other person can see the real me.
What worries me most about AI-generated content is this: it can compress the complexity of being human into standardized expression. It can turn real emotion into "appropriate" scripts. It can rewrite the uncertainty and vulnerability in relationships into something that looks like a perfect way to communicate.
So I don't want to reject AI. I want to set boundaries for it. Let it help me polish my sentences — but not replace my own voice. Let it fix my grammar — but not decide what my emotions should look like. AI should not get to decide who I am, how I feel, or how I choose to express myself.
I hope more people bring a social work perspective into conversations about AI. When a technology this powerful enters our daily lives, it shouldn't only be discussed in technical terms. We also need to look at it through the lens of humanity — the boundary between humans and machines. How does AI change the way we understand ourselves, express ourselves, and build relationships with others?
Responsible use of AI isn't about chasing endless fluency or efficiency. It's about keeping the person at the center. Crying, pausing, clumsy words, grammar mistakes, imperfect expression — these are not always problems that need fixing. Sometimes, they are proof that a real person is still there. As AI changes our world so quickly, we still need to hold on to our human agency.