Metacritic doesn't review games directly. Rather, they algorithmically compile, normalize and weight reviews from industry publications. Their trustworthiness is dependent on how well their algorithm reflects the overall opinion of the gaming press. I certainly won't try to perform a systematic analysis of MC's scores, but I will say that in the case of GTA IV, they're spot-on. GTA IV broke all sorts of sales records and was named by most of the top publications as the best, or one of the best, Xbox 360 games to date. Metacritic's score accurately captures the industry opinion, circa 2008.

...but not necessarily accurate. The industry as a whole might be biased, selling out to the publishers, or just flat out wrong. Trouble is, they're the best we've got. User scores (you mention that you've heard lots of people complain about GTA IV) are even less trustworthy than the gaming press. Sort Metacritic by user score and tell me what you find. Do you think, by any objective criteria, "Guwange, "Fight Night Round 3," and "Culdcept" are the best 360 games of all time? Branching out into PC, do you think Diablo 3 rates a 3.8, or Simcity a 2.0? User scores are impacted by everything from gameplay to business practices to Twitter comments. They do not always, or even usually, reflect a game's quality.

Anyway. Metacritic is far from perfect, but it's less susceptible to bias than any single industry publication, and is a lot more accurate than listening to random internet commenters. Finding a particular reviewer or friend whose opinions you usually agree with is probably the best way to get gaming recommendations, but barring such personalized searches, Metacritic is probably the way to go.