I doubt it's "as advanced" as doing keyword matches on the page in question, though they might well do some of that as well. I suspect the main reason for hitting the link is to resolve redirects.

I run a plain Mastodon install on https://m.galaxybound.com/ and I can post links to it just fine. If it can't even detect a standard Mastodon install, it's not a very successful search for blocked content.

It seems to me they're stupidly still maintaining a blacklist of the larger instances.

Note that I've "even" tweeted links to a post on my Mastodon instance that contained links to a blocked instance, and Twitter didn't even detect that.