DeepL at one point allowed user correction, doesn't seem like they have it now. But the longer the service has existed and been "corrected", it's gotten both better and worse with reading Japanese. It can handle half-width kana, but if you don't give it a period at the end of the sentence it could distort the whole response it gives.
They can't be worse, but they could be just as bad, if the AI itself has been suborned with the same agenda.
Looks at Gemini
Or through user input nudging
DeepL at one point allowed user correction, doesn't seem like they have it now. But the longer the service has existed and been "corrected", it's gotten both better and worse with reading Japanese. It can handle half-width kana, but if you don't give it a period at the end of the sentence it could distort the whole response it gives.