An individual molecule of H2O doesn’t have any of the observable properties we associate with water. A glass of water, pure as water can be, is better understood as containing H2O, OH–, H3O+ and other related but less common ions, and even this is a vast oversimplification (if we could get truly pure water, which we cannot). Our current best understanding of the electron transfers that give water the properties we observe is a statistical average of ever changing interactions so complex as to be quite literally unthinkable. Indeed, the problem is “not that we are unsure which (distribution of types of) microstructure is the correct one. The point is that there is no one correct microstructure, because the microstructure depends as much on the context and functions just as another nominal essence would” (van Brakel, 2000b, 80–81).
This is why chemists use the ‘mixture of ions’ model to describe water’s macroscopic behavior. The only thing we can say about a glass of water that is not, strictly speaking, an error is that the average ratio of atoms in the glass is 2 H: 1 O and that it has the macroscopic properties of water. If there are other kinds of atoms in the glass, or if the ratio is other than , then we do not have pure water. If the ratio is but it does not have the macroscopic properties (pH, boiling point, etc.) of water then we have not water but a mixture of hydrogen and oxygen molecules in their elemental form. Chemical analysis and experiment deals with water not as an individuated molecule but in macroscopic quantities. All the typical observable properties of water—its pH, its density, its boiling and freezing points, its utility as a solvent, are dependent not upon its atomic ratio but the interactions between the dissociated ions. Philosophers of chemistry have been arguing this point for at least 25 years....
The idea that water simply is H2O is one of those false reductions that people can't seem to get out of their heads. What is actually true is that if you break apart the bonds in water you'll get a ratio of hydrogen to oxygen that's roughly 2:1; a very large amount of this will be from H2O, some from OH, some from H3O, etc., and it is the overall interaction of all of these, not just properties peculiar to H2O, that give us what we call water, because the properties of water arise not from the molecule but from constant, widespread dissolutions and reformations of H2O and related molecules. The matter is complicated by the fact that not even all H2O is the same; most involves the protium isotope of hydrogen (one proton), a small number have deuterium (when you have water in which a very high percentage of the hydrogen is deuterium you have heavy water, which is used in nuclear reactors -- heavy water is water, but it's poisonous in large quantities), and a small number involve tritium (which in large quantities would give you tritiated or super-heavy water, which is corrosive and, if I am not misremembering, radioactive). And that's not even counting the fact that the oxygen can occur in isotopes 17, 18, and 19, each one resulting in molecules that behave differently. Water is an interacting society, not a molecule, and it is a society of related molecules, not just H2O; among those molecules H2O is just the most prominent family, not a single kind of molecule; and light water (what we usually think of as H2O) is just the most prominent branch of that family. What looked like a simple fact is in fact not simple at all.
ADDED LATER (September 2018): Since this continues to be a fairly popular page, a few updates are perhaps in order. The following is a good summary discussion from a chemist's perspective of the (many, many) complexities that are involved in understanding water:
Martin Chaplin, Water Molecule Structure
Michael Wesiberg's paper is also available at Academia.edu. It's still worth reading.
If you have JSTOR access through a library, you can read van Brakel's paper there. VandeWall's paper is also available there.
B. C. Malt has an older paper, "Water Is Not H2O" [Cognitive Psychology 27:41--70 (1994)], that is written from the perspective of cognitive science, and does a good job arguing that people in ordinary everyday life also don't necessarily think of 'water' as designating a single kind of thing (e.g., there is a functional component). This is relevant because I think one reason why the water-is-H2O reduction won't die is that philosophers radically underestimate how complicated the requirements for good classification are when it comes to water -- and this is true for ordinary language as well as for chemists. You can read the abstract for Malt's paper here.