This is a documented restriction which punishes users for using the basic feature of underlining part of a word. Given the current state of software usability, it is no less than a bug.
apps.ankiweb.net states “Anki supports complex workflows, but getting started is easy.” Bold/italic/underlined text is basic formatting and not advanced. The ability to find text as it is displayed is basic and not advanced. Getting search to work is not easy.
I assume the solution involves extracting the text from card sides and caching it for search in-app. Hopefully this would also help with unicode letters.
This was reported as a bug multiple times (one, two, three). This post is not a bug report. This is not a request for extensions which do not work on Android anyways.
I read it and you just complained about it not being possible in a long winded way. Nothing about it suggests that you understood that it is possible, albeit a bit unintuitive.
The “$:text” search type is now supported. This will find all words, even if each character is followed by an diacritic (a mark that changes the sound) or if there are tags between letters (bold, etc.) – it will find everything.
The search is not case-sensitive.
To perform this search within a regular expression ("re:"), you must write it as a call to the $(text) function.
How it works: instead of $:abc, the search will be "re:a[\u0300-\u036F]*(<[^>]*>)*"b[\u0300-\u036F]*(<[^>]*>)*c"
