It’s a fine line between relying on graphics vs explicitly labeling things. In general Discourse needs to provide some more help in my opinion in making on-ramps for early and non-technical users. Perhaps this isn’t that place, but I think it is.
I really couldn’t tell you without looking what the visual distinctive of a tag was. In fact, it appears to be the lack of any distinctive that is the one distinctive of tags. In a context like this, that reasoning by a process of elimination is expecting quite a lot of users.
Edited to add: regardless this is a nice change to search.