I'm checking how good/bad is the Google AI to answer some grammar questions. In relation to the "て+いる" form it says:
1. Habitual Present (Everyday Actions / Routines) -> ます Form VerbsWhen you use words like 毎日 (mainichi / every day), よく (yoku / often), or いつも (itsumo / always) to describe a normal routine or a general habit, N5 requires using the simple present (ます form).私は毎日映画を見ます (I watch movies every day -> This is my routine/habit).毎日学校へ行きます (I go to school every day).
2. Continuous State or Action in Progress -> ている (ています) FormIn the beginner level, the V-て + いる structure is reserved mainly for two things:Actions happening right now (equivalent to the English "-ing"): Ima eiga wo mite imasu (I am watching a movie right now).Resulting states that last over time (such as wearing clothes or marital status): Kekkon shite imasu (I am married), Megane wo kakete imasu (I am wearing glasses).
It is true that in natural Japanese (from N4 level onwards), the ~te imasu form can also be used with 毎日 (mainichi) to emphasize that a habit started in the past and you are still actively doing it today. However, for the N5 level, if the sentence describes a simple routine with 毎日, the correct answer will always be the ます form.
How reliable is this information? I'm learning towards the N5 and I find this point important.
At first glance, if you strip away the weird JLPT-level commentary, the grammar description itself seems accurate enough for a beginner. A general explanation of ている is something I'd expect AI (depending on the AI that is) to do pretty well on. Although what you got is uncharacteristically messy. I'd expect something much cleaner / easier to digest.
I personally still wouldn't recommend relying on it alone, since you can easily find reliable articles/videos for ている. Like this one for example: https://www.tofugu.com/japanes...
Keep in mind that AI is weak for nuance, edge cases, and usage preference. I've seen it directly try to contradict nuances that natives generally seem to agree on. That's not really an issue at a beginner level though.
Also, I haven't checked in some time, but I remember ChatGPT being absolutely terrible at distinguishing between JLPT-levels. You'd ask it for a practice text with kanji up to N4, and it would mix in a bunch of N2-N1 kanji.
Edit: I tested it and it's actually much better than it was 3-4 years ago. So I guess it's ok for generating some quick reading practice material. I'd still recommend graded readers over it though :D
If you use (insert AI of choice), you're going to have to check every single thing that comes out of it. When it's wrong, unlike humans, there will be no hedging language, "I think", or anything else that we use to assess confidence in someone's language.
ている generally has three usages, not two (all three are mentioned in renshuu's lesson, found under Resources > Japanese Lessons > Japanese Basics - we even have a video where I break it down in more detail).
The part that talks about "resulting state" is way too simple, to be honest. "Resulting state" is true, but it really needs to describe what state is (and that it actually is not state, but state resulting from a specific, discrete change) to get a good feel for it.
Some people use renshuu, others use tofugu, but there are tons of grammar resources out there that were written way before AI, and since grammar doesn't *really* change, stuff written 10+ years ago is still quite relevant today.
