Can ChatGPT Teach You English? What AI Tutors Get Right (and Wrong)

Ask ChatGPT to explain the difference between "since" and "for," or to roleplay a job interview in English, and it will do a genuinely good job. Ask it to make sure you still remember that explanation in three weeks, and it can't — not because the model is weak, but because that isn't the kind of problem a conversation can solve. This post is an honest look at where AI chat tools actually help English learners, where they fall short, and why the gap between the two is a memory problem, not a knowledge problem.

TL;DR:

• ChatGPT is genuinely good at instant grammar explanations, judgment-free conversation practice, and generating example sentences on demand.
• It has no built-in review schedule and no memory of what you've already learned — a good explanation today doesn't come back to you next week.
• Spaced repetition and active recall solve the "sticking" problem chat tools aren't built for; use AI to explain and practice, use flashcards to remember.

What ChatGPT is genuinely good at

A few things AI chat assistants do well for English learners, without much hedging:

• On-demand grammar explanations. You hit a sentence that doesn't parse, and instead of digging through a grammar site, you paste it in and ask why. The explanation is patient, specific to your exact example, and available at 2am. That's a real improvement over a static textbook.
• A conversation partner with infinite patience. Practicing spoken or written English with another person means scheduling, and it means feeling embarrassed about mistakes. An AI chat doesn't get bored, doesn't judge, and will happily repeat a roleplay scenario five times until you're comfortable with it.
• Instant translation and paraphrasing. Not knowing the word for something used to be a full stop in a conversation. Now it's a two-second detour.
• Example sentences on any topic, on demand. Want to see "nevertheless" used in a sentence about cooking? About sports? About your actual job? An AI chat can generate as many variations as you want, tailored to contexts a textbook would never think to cover.
• Judgment-free repetition. Asking the same question five different ways until it clicks is something a lot of learners are too self-conscious to do with a human teacher. An AI doesn't care.

These aren't small conveniences — they genuinely remove friction that used to stop people from getting an explanation or a practice rep when they needed it, not hours later.

What ChatGPT isn't good at — and why

It has no memory of what you've already learned. Open a new chat tomorrow and, by default, it doesn't know you struggled with "since vs. for" yesterday, or that you already asked about phrasal verbs with "up" twice this week. Some tools now offer persistent memory features, but they're opt-in, easy to forget to manage, and not designed around your vocabulary specifically — they remember scattered facts about you, not a structured record of what you know cold versus what you're shaky on.

There's no review schedule. This is the core issue. A good explanation in a chat window is a single exposure. Nothing brings that word or rule back to you tomorrow, or in four days, or in two weeks — the exact timing that would need to happen for it to move from "I understood it when I read it" to "I can produce it without thinking." Left alone, new information doesn't just sit there waiting to be recalled later — it decays, and fast. That's the forgetting curve: most of what you learn in a single sitting fades within a day or two unless something brings it back at the right moment. A chat conversation has no mechanism to do that bringing-back. It's not designed to.

It can't track your actual progress. A conversation feels like progress in the moment — you understood the explanation, you used the word in a sentence, it felt good. But "felt good in the conversation" and "will still know this next month" are different claims, and only one of them is something an AI chat is set up to measure or influence.

Pronunciation and speaking feedback is inconsistent. Voice-mode AI tools have improved a lot, but feedback on accent, stress, and intonation is still noticeably rougher than feedback on grammar or vocabulary. It'll catch an obviously wrong word. It's much less reliable at "your stress is on the wrong syllable in a way that changes the meaning."

None of this is a knock on the technology doing something wrong. It's a mismatch between what a chat interface is built for (generating a good response right now) and what long-term retention requires (a specific word or rule coming back to you at specific, increasing intervals, whether or not you happen to open that app that day).

The honest insight: explaining vs. sticking are different problems

AI chat tools are excellent at explaining and practicing in the moment. They are, as currently built, not good at making anything stick over weeks and months. That's not a small caveat — it's the entire difference between "I understood this once" and "I actually know this now." Understanding something is necessary but not sufficient; the thing that turns a good explanation into a permanent piece of your vocabulary is what happens after the explanation, and a chat window has nothing built in for after.

This is where spaced repetition comes in — not as a competitor to AI conversation practice, but as the piece it's missing. Spaced repetition is a scheduling system: it tracks each word or fact individually and brings it back right before you'd naturally forget it, so a single good explanation turns into something durable instead of something that faded by Thursday. It solves a completely different problem than a chat assistant does, which is exactly why the two work well together rather than against each other.

There's a second half to this, too. Even a perfectly-timed review only works if you're doing the right kind of review. Active recall — trying to produce an answer from memory before checking it — is what actually strengthens a memory. Rereading a ChatGPT explanation a second time feels productive, but it mostly builds recognition, not recall; you'd recognize the word if you saw it again, which is a much lower bar than being able to produce it yourself, unprompted, in an actual sentence. A flashcard that makes you retrieve the answer before you see it is doing something a re-read of a chat transcript isn't.

Use both — they're not the same tool

Put concretely: ask an AI chat assistant to explain a grammar point you're stuck on, or to have a five-minute practice conversation in English about your weekend. That's a genuinely good use of the technology, today, as it exists. Then take the vocabulary and phrases that came out of that conversation and put them somewhere with an actual review schedule — something that will bring "since vs. for" back to you in a day, then a week, then a month, instead of leaving it to chance.

That's the gap Budling's flashcard library and placement quiz are built around — not a replacement for AI conversation practice, but the memory layer underneath it. Chat with an AI to get unstuck and to practice in the moment; use spaced repetition to make sure what you practiced is still there next month. They're solving different problems, and neither one is a substitute for the other.