GR Stories | What AI Actually Can't Do When Studying in Another Language
Let me start with something that will sound obvious once I say it, but that took me longer than I expected to actually notice.
AI is very good at sounding like it understands you. And that is not the same thing as actually understanding you.
This distinction matters everywhere. But it matters in a particular way when you are an international student, studying in a language that is not your own, trying to use AI tools to help you get through an academic year in a foreign country. Because in that situation, the gap between "sounds like it understands" and "actually does" can cost you more than you realise.
I am not writing this to tell you to stop using AI. I use it myself. Most students I know do. And there are things it genuinely does well — well enough that ignoring them would be its own kind of mistake.
But there are limits. And those limits tend to cluster around exactly the things international students need most.
What AI is actually good at
Before getting into the problems, it is worth being honest about what works.
If you are studying in a country whose language is not your first, AI can be a useful first layer of support. When you encounter an academic text that uses vocabulary you do not know — discipline-specific terms, legal language, dense administrative writing — AI can give you a clear explanation. Not always a perfect one, but a workable starting point.
It can also help you draft. If you are writing an essay or a report in your second language and you know what you want to say but you are uncertain whether you have said it correctly, running a draft through an AI tool can catch obvious errors and make the language sound more natural. This is genuinely useful. It is not cheating any more than using a dictionary is cheating.
For translation, AI is faster and generally more accurate than most human quick-translation options, and better than many dedicated translation apps. Not perfect — and the imperfections matter, which I will come back to — but faster and cheaper than alternatives.
And for summarising long documents, AI can save real time. Lecture notes, research papers, university regulations, housing contracts: AI can condense all of these into something more manageable, which is useful when you are operating in a language that makes reading slower and more effortful than it would be at home.
So yes. It helps. In real, practical ways.
Where it starts to quietly fail you
The problems begin not in the obvious places — the catastrophic, clearly wrong answers that you immediately recognise — but in the subtle ones. The answers that sound right, that have the right shape and the right level of confidence, but that are slightly off in ways that are hard to catch unless you already know enough to catch them.
This is the particular trap for students studying in a second language. Your ability to verify what AI tells you is limited by your own fluency. If an AI tool translates something incorrectly or explains a concept in a way that sounds plausible but is not quite right, you may not have the linguistic or contextual resources to notice. You accept it. You build on it. And sometimes the error compounds.
There is a specific version of this in academic translation. AI translation is, as I said, generally good. But "generally good" is not the same as "reliable for academic purposes." Academic writing in any discipline carries a specific register — a way of using language that is precise in ways that are invisible unless you have spent time inside that field in that language. Legal texts, philosophy, economics, engineering — all of them use ordinary words in technically specific ways that AI translation regularly misses or flattens.
I have seen this happen with Italian administrative documents. The word "decadenza" in Italian administrative law does not simply mean "decline" or "decay." It means a specific kind of forfeiture of a right due to inaction within a prescribed period. An AI might translate it accurately in some contexts and inadequately in others, and if you are a non-Italian student trying to understand whether you have lost your right to appeal something at your university, that distinction is not a small one.
The problem with using AI to write in a language you are learning
This one is more personal, and I feel more strongly about it.
Using AI to help correct your writing is reasonable. Using AI to write your academic work for you is a separate category of problem — not just ethically, but practically, and in a way that is specific to international students.
If you are studying in your second language, then writing in that language is part of what you came here to do. Every essay you struggle through, every sentence you construct badly and then revise, every paragraph where you realise you do not know the word you need and have to find it — all of that is building something. It is building your fluency, your academic vocabulary, your ability to think in the register that your discipline requires in that language.
If you outsource that to AI, you are not cheating only in the institutional sense. You are cheating yourself out of the thing that will actually make your year here worth what it costs. You will arrive at your final exams, or at a job interview conducted in that language, or at a professional situation that requires you to write formally in it, and you will realise that the fluency you thought you were developing was being quietly replaced by a tool that will not be available to you in that moment.
This is not a moral argument. It is a practical one. The cost of using AI to write your second-language academic work is that you do not learn to write your second-language academic work. And that skill is not optional. It is the whole point.
AI does not understand your actual situation
There is something else that is harder to articulate but worth saying.
When you ask an AI tool for help as an international student — with a university process, with understanding a bureaucratic requirement, with navigating a system you are unfamiliar with — the answer you get back is a generalisation. It is based on patterns in data. It has no idea which university you attend, what year you are in, what your specific programme requires, what the actual rule is at your institution right now, or what exceptions might apply to your case.
It sounds like it knows. The confidence of the prose suggests that it has specific knowledge of your situation. It does not.
I have found this most sharply with visa and permit questions. Students ask AI tools about student permit renewals, about what documents they need, about whether a particular administrative step is required. AI answers these questions in a way that sounds authoritative. But permit regulations differ by country, by municipality, by year, sometimes by the specific university you attend. An answer that was correct six months ago may not be correct now. An answer that applies to students at one institution may not apply to students at another.
This is exactly the kind of question where the cost of a wrong answer is high. Getting visa-related information slightly wrong is not a small mistake. It is the kind of mistake that can affect your legal status in a country.
For questions like these — and there are many versions of them that come up in international student life — AI is not a substitute for the actual office, the actual website of the actual authority, or the actual human at your university's international student services desk.
The emotional limits
There is one more limit that I think gets overlooked almost entirely in conversations about AI and student life, and it is the most human one.
International students deal with a specific kind of stress that has no straightforward practical solution. The stress of being far from home. Of not quite knowing yet whether this was the right decision. Of feeling more alone than you expected, in moments you did not predict. Of watching other people seem to settle in faster, or build friendships more easily, or navigate the system with less visible effort.
AI can produce responses to these feelings. It can generate something that looks like empathy. It can offer suggestions. It can say the right kind of words in the right kind of order.
But it cannot actually be present with you. It cannot sit across a table from you and understand — not from a database of descriptions of how loneliness works, but from having felt something like it — what you are going through. It cannot follow up the next day to see how you are doing. It cannot make you feel less alone in the way that another person can, even a person you barely know yet.
There is a risk, I think, of using AI tools as a substitute for the human connection that is harder to build but more necessary. It is easier to type into a chat interface than to send an awkward first message to a classmate. It is easier to ask an AI to explain your feelings back to you than to call someone from home. But easier is not the same as better, and in this particular area, it is not even close.
How to use it well
None of this means you should not use AI. It means you should use it for the right things and with the right expectations.
Use it to clarify vocabulary and explain concepts you do not understand — but verify anything academically important through other sources. Use it to catch errors in your writing — but write the writing yourself. Use it to get a first approximation of how an administrative process works — but confirm the details through official channels. Use it as a starting point, not a finishing point.
The students who get the most out of AI tools during a year abroad are not the ones who use them the most. They are the ones who are clearest about what AI is actually doing — which is making a very good guess, very quickly, based on patterns — and who use that for what it is good at, while not expecting it to replace the things that require real understanding, real presence, or real effort.
Real understanding of a new language and a new academic system.
Real presence in the friendships and communities that make a year abroad worth living.
Real effort on the work that is supposed to be hard.
Those things remain yours to do.
