What you do with the clicker depends on how you're using it. For example:
Are you free-shaping? In a nutshell, that's basically playing hot-and-cold with the dog and creating a behavior by c/t (click/treating) approximations of the finished behavior. For instance, for a sit, you might c/t when you see the dog start to offer a lowering of his butt to the floor (whether or not he actually sits is irrelevant at this point). Do that several times 'til you see the dog start to connect the fact that lowering his butt toward the floor is making you c/t. Then raise your criteria. Decide you want the dog to move his butt closer to the floor and ONLY c/t the closer responses. The dog will likely do the first activity (the one that initially got him paid) several times thinking, "Hey! This used to work... what's up?" Then, out of frustration, they try just a little harder, and that extra effort = butt closer to the floor. BINGO! You c/t that. And for the next however many reps (until you see the dog consistently offer THAT version of butt-toward-floor, you ONLY c/t that version. Raise the criteria again, maybe to butt-ON-floor, etc. That's shaping.
You can also jump-start a behavior by luring a handful of times (say, no more than 5) and still c/t when you get the behavior. So, for the sit, you use cookie on nose, raise over dog's head to lure into a sit and then c/t as the butt hits the floor. After about 5 lure trials, just sit and be patient and see if the dog "offers" a sit... BINGO! c/t... and continue.
There's really no right way to "clicker train." The "purists" will say it should be about shaping, but other segments of the reward based (R+) population will say that it's also a strong marker signal that is far more salient to the dog than a verbal marker, which allows it to be beneficial even in lure-reward training (the cookie-on-nose technique).
In general, it's best not to name a behavior in the early stages. Let your training (shaping or luring) make the behavior happen and when you see the dog start to reliably offer it and you can predict that the dog's just about to do it, then you slot the name in a nano-second before. If you start naming it too early, the dog doesn't know it yet, so the words potentially get associated with the unfinished version of the behavior.
A really good resource is
www.clickertraining.com.