Perhaps the reason that a majority of scores from online IQ tests are above average is that such tests have a self-selection bias toward those with above-average intelligences.
Most seem to realise that online opinion polls will have self-selection biases, and so it should not be very surprising that websites which deign to bestow a numerical value to one's intelligence will also tend to draw certain types of people disproportionate to their incidence in the general population - such as the keen and the curious.
As for the concept of IQ tests, they are crude at best. Since intelligence is modular in ways more sophisticated than we can even pretend to know, sincere IQ tests must attempt to probe a common denominator between them. Such a denominator is speculative, to put it mildly, and naturally precludes a capacity to ferret out the peculiar nuances which distinguish the great geniuses from each other.
And being able to notice and accurately identify such distinctions is not optional when attempting to seriously classify human intelligence at its highest levels. This inevitably condemns single-numbered portrayals of brilliance as absurdly simplistic imposters.
This is not, of course, to imply that such tests are entirely worthless. They are perhaps most useful (aside from entertainment value) when wishing to ascertain certain relative skills capacities in children. But childhood scores bequeath no practical relevance in adulthood. Prodigies are notorious for fading into obscurity upon coming of age.
If there were no IQ tests, how would one go about estimating the intelligence of another? We rarely know anyone else's supposed IQ, and so we are forced to infer by other means all the time. Is it a useful activity to speculate whether Leonardo was smarter than Bach? Not as useful as other things we can spend our time on - and our time, after all, is very limited and precious indeed.
