I assume the request for clarification of "Plain-Text" has to do with content, not that it's an ASCII or UTF string.
There is not a "Hello World!" string that I could make that you couldn't then copy and call your own, where I could 100% prove that it was mine to begin with. For example anyone can just delete a PGP signature wrapper.
However, if the content of your string is structured in a way that was uniquely identifiably yours then I would call that a watermark, as any change to the content of string would change its context. For example using a private RSA key to encrypt your string then coding it back into ASCII with Ascii85, or if your string contains obfuscated JavaScript, or even simply something you've published under a copy right
That should afford about the same level of watermark you'd see in the other media, but the thing is any media which can be displayed with out the watermark is by the reflexive property media that can be stripped of its watermark and not still remain self judicating. ie copy right it, or have context that's destroyed when changed.