Beyond usual domains, if the link to a site is something.onion, you can access it via the Tor Browser Bundle. If it is something.i2p, you need to download and run the i2p bundle; similarly for Freenet, GNUnet, et cetera. But the real question is how you get the link in the first place; as Chronos explains, if a site shows up on a search engine or a wiki, it is ipso facto not a part of the deep, dark web, no matter how many technical hoops one must jump through in order to access it.
Because there are what TOR calls hidden services, or machines you can only access via the TOR protocol. These include websites, email servers, IRC chat, and other services.
These services are hidden because it’s impossible to figure out where in the world they are just by accessing them over TOR.