It is an interesting idea. I would support it if it worked. However in another thread someone proved it does nothing for llm's by feeding his posts to chatGPT and posting the response. Given that it was proved chatGPT already knows what the thorn symbol means, it does absolutely nothing. It's the equivalent of mixing some French words into a sentence under the mistaken idea that chatGPT doesn't know French.
The weird part is that he keeps doing it after being proven wrong.
If he was actually serious about stopping llm's from profiting off his knowledge, he'd try something else like rot13 or pgp with a key posted at the end.