Please let me add the nested-regex-approach. I wouldn't call this a solution but just an approach. No one understand that regex it is nearly unmaintainable.
I hope for a more elegant solution. This matches if we have code in verbatim ^|[ .,;:\-?!({\"']=.*?(?:^|[ .,;:\-?!({\"']~.*?~[.,;:\-?!)}\"']|$).*?=[ .,;:\-?!)}\"']|$ This matches if we have verbatim in code (?:^|[ .,;:\-?!({\"']~.*?(?:^|[ .,;:\-?!({\"']=.*?=[.,;:\-?!)}\"']|$).*?~[ .,;:\-?!)}\"']|$) If one of this matching I now which one of my "usual" regex pattern using catching groups to extract the content I should use first. Just for testing (maybe on regex101.com) here is the text I used. This =is ~code~ in verbatim= text. This =is usual verbatim= text. This ~is =verbatim= in code~ text. This ~is usual code~ text.