The danger here is that you're not learning German - you're learning machine German. Even if the app makers have structured the machine translation so that it's smart about taking context into account, it will be at best subtly different from actual German, and at worst you'll pick up nonsense and think it's good German.

The safe way to use this would be in reverse. You shouldn't be browsing English pages and get 1 in 10 translated into German. You should be browsing German pages and get 9/10 translated into English. You'll still get machine translation artifacts, but they're much less likely to interfere with your learning, and you'll be much better equipped to spot them.

For sure, some common ground is needed. Madrigal's is a famous example of LT that almost exclusively focuses on transfer itself, because English and Spanish are reasonably similar.

LT's method goes a bit further though, hence why there are courses on Arabic, Swahili and (upcoming) Japanese. Chinese might be even further removed, but the LT courses are about learning to _think_ about how the language works, and the format of the course (teacher + median student + you) goes a long way to encourage this. Beyond Madrigal's "look how similar these words are".

Its in the article itself.... Nothing wrong with coming straight for the comments but this is what the author recommends as an alternative lol