There are two contexts in which you can ask and answer this question - the strategic and the tactical.
Adapting a campaign somewhat at the high level to your players is generally considered good form. At the point before play starts, you always have a lot of work to define what will be in the world and go on. The argument "the world is fixed" is not legitimate at this point, because no RPG world product I have ever bought actually describes every person and event in the world. If one player has diplomatic skills, putting in a merchant for him to negotiate with is not "breaking" the world. If a player makes an "undead killer" character, and your world has undead, and you never bring any across his path, he'll be sad and you'll be a dick. The GM is the motive force in the world and in the end it's an arbitrary decision on your part whether the zombies in the graveyard are feeling frisky or not, so appealing to the objective world if you say they're not is inappropriate.
Furthermore, world building is almost always an exercise in horizon planning. You don't build the entire thing down to the detail to which the PCs will interact with it at start, you end up detailing new villes or whatever as you go. Unless you have the world's most complete random generation table, you are making decisions as to what goes in there, and how that stuff will interact with your players is an important factor in everyone's fun.
This can, however, be taken too far in terms of catering to the players. In all games except the most simplistic, a character with diplomatic skills can exercise them by wandering up to someone and, you know, talking to them. When GMs design TOO much around the characters, it "blunts their edge" - they become passive, and instead of looking for people to talk to or going around and looking for undead-haunted crypts to loot, they just kick back and expect you to line up "appropriate challenges" for them like tenpins. There's a reason why people that get good at other games enjoy those games more - the practice isn't always fun in and of itself but the play experience becomes so. Whenever my daughter starts a new sport she hates it until she starts to get good at it and then she likes it. Letting players not try undermines the fun long term. It forces interesting choices on them - "Do I use a longsword, which I'm good at, or this magical axe, which I'm not as good at?" It can force them outside their comfort zone.
A lot of it has to do with what your players like and what kind of story you want to tell with a given campaign. In one campaign, the GM had all the players be growing into avatars of the gods, so naturally a lot of the game world revolved around the PCs - but not at a metagame level, at a quite-reasonably-in-game level.
Some GMs will adapt the game while in play to play to (or against!) a character's strengths. I remember fondly one very realistic campaign I played in. The GM would, for example, detail out a bad guy's manor house, and we would plan a careful intrusion, because we were thieves. If we chose the passageway without the guard, he would not "move the guard" so that we'd come across him anyway. This rewarded our investment in the game world and in planning. I generally did not play thieves, because most GMs I had played under would do that - therefore your cleverness was without real merit; if the adventure said "you have to fight the guard to get in" then you had to fight the guard to get in. I found that terribly demotivating. One can argue that it preserves GM prep work, as in "they spent a lot of time on that guard so it's a shame to not use them," or that "the challenge won't be complete if you don't have a fight," but both of those are pretty transparently nonvaluable.
This level of "adaptation" is the one usually perceived by players as cheating/fudging/railroading/contrived, whether the GM thinks it's 'for' or 'against' them. "Oh, they don't have a lot of spells left, this guy will take it easy on them." "That darn PC with his bow attack, I'll give every batch of bad guys some ranged defense thing." This can be good if you value a standardized challenge for every encounter, or if you really want the story to go a specific way regardless of the will of the characters, but I find that in general players don't value those things.
My Personal Mix
I have a larger world that "exists independent" of the PCs. I make a campaign plot that takes them into account in a large-scale sense. When I plan a specific session, I look to have things show up/happen that should be relevant to characters, maintaining a mix of tailored/random/absolute genesis. Then when I start the session, it's 100% simulation till the session is over with. It provides the right amount of each of the good things to keep players coming back consistently for 20 years without falling victim to too much of the bad things.