If the King's Gambit is unsound, how could a world champion play it often?
The King's Gambit has been "refuted" at several points in the past. Most have later been proven incorrect, IIRC only two lines at this point seem suspect. But that's not the point. The point is: how many of your opponents know enough about those lines to play them as accurately as required to win? (John Shaw in his book on the KG noted analyzing one of the resulting positions overtaxed the water-cooling in the computer he was using to help analyze it.) Refuting something in theory does not equate to refuting it in over-the-board play.
Bent Larsen once said "I don't often play a move I know how to refute." Which, of course, means sometimes he does.
"Refutations" of white openings often fall into the category described above, which is that white fails to maintain the advantage of the first move. Give that some thought while asking yourself the question: "would I rather play a position that retains a slight advantage if I follow a fixed and complex series of moves, or would I rather play a level position that I understand completely and am comfortable playing?"
It's that last bit that tips the balance. I know a national master that loves to play the Blackmar-Diemer Gambit. His preference is so well known that everyone he plays has the opportunity to "book up" on that opening, of which by most reports "fails to maintain the white opening advantage" would be a kind description. Yet he won. Consistently. Yes, even against other masters. (I've won more than I've lost with it myself, though I don't make a "steady diet" of it.) He was so familiar with the strategies and tactics of the resulting positions he was comfortable playing them against anyone.
And that's the point. If it's a position you're comfortable playing, it doesn't matter even if a computer says you stand a few centi-pawns worse. (Heck, unless your opponents are routinely over 2200ELO even a theoretical full pawn difference may not matter.) If you understand and are comfortable in the middlegames that arise from it, you'll win more than your "objectively evaluated" share of them.
Spassky did not play it "often" and you'll notice he also didn't play it in any of his 3 world championship matches. http://www.chessgames.com/perl/chessplayer?pid=21136
Black can very quickly equalize, for example 2...Bc5 or 2...exf4 3.Nf3 d5. Among many others. Black can also try to maintain his opening advantage with lines like 3.Nf3 d6, among others. In fact black has so many options to be well prepared as white you need to know nearly a dozen different primary variations.
So these are the factors that make it impractical:
1) Black can equalize easily (professionals usually want more with the white pieces).
2) It requires more preparation than other openings (this is troublesome for amateurs who don't have all day to work on their openings).
3) Black is in the driver's seat regarding the type of position that will be played (stale or dynamic).
The King's Gambit has been analyzed to the point, where White fails to retain his advantage with proper play by Black. That's why it is inadvisable to use it against opponents as strong as you.
But it is an aggressive opening that works well against weaker opponents. Spassky may have won more games against weaker players using it than players of equal strength playing different openings against the same opponents. Spassky's record was among the "best," until he ran into top-flight competition, notably Fischer.
That is all very silly. It the KG is as fine in most lines for top players as much as for amateurs. The computer gives equal or tiny negative in most lines, plus in others. But does so when white is down a pawn. That means that it considers that White has an attack advantage that is equal to a pawn. Equal from a computer when you have gambited a pawn means it is a good line. If you read the rubbish people write on these forums you would think that if you played it against a gm over 2500 that you just lose by force. That isn't true at all, there are many wins and heaps of draws against super GMs in the last 30 years. Yes if you go into some terrible line in the fischer defence you can get a horrid position where you are fighting for a draw, but there are many options for white that avoid this. And if you look at super GM games in last 30 years with the KG, you will see that almost none of the GMs play the so called critical lines from black, likely because they are only super critical if you are a computer or are playing correspondence chess.