For all that I'm not totally sold on this article's idea of "stems" and "suffixes", I think it does a good job of avoiding this pitfall and correctly explaining the groups.
An important aspect of vocabulary that informs you about verbs is knowing the nominal: the noun-like stem. Like when you consider 帰る, you know from your vocabulary that there is no noun-like word "kae" that is just written 帰. You know that "homecoming" is not "kae" but "kaeri", 帰り.
So from that alone you know instinctively which way it conjugates: らない、れない、りたい、った、って、ろう、。。。
And a big source of learning the stem is ... japanese polite speech with -masu.
This is how children absorb what the stems are: hearing all the verbs in -masu form.
For instance, I think that every single verb that ends in "-rimasu" in the polite form (other than just arimasu?) is a godan -ru word. Drop the masu and you have a -ri stem, which implies the word is -ru, and conjugates godan: -ranai, -renai, -ritai, -rō, -tta, -tte, ...
Not quite: https://jisho.org/search/%23words%20%23v1%20%3F*%E3%82%8A%E3...
None of these are ever other than ichidan? If we imagine a -riru verb that is godan, it would have to go to -ririmasu.
That sounds like something only a ditzy weather girl called Ririko would say, for a gag. :)
FWIF, the "#words #v5r ?*りる" search does come up two hits:
The first one is 手を借りる, which is a classification error; it is a compound verb that conjugates exactly like 借りる by itself.
The second one is a curiosity: ラリる -- to get high on drugs. Supposedly godan, so it does give us ラリります, etc.
Wikipedia page on it with some etymology/origin notes: https://ja.wikipedia.org/wiki/ラリる
Well, apologies to Ririko there!