Realised volatility is the standard deviation of returns over a past window, annualised. It is a measurement, and it is the closest thing to a fact in the whole volatility discussion. It also involves several choices, each of which changes the answer.
The window length is the largest choice. A 10-day realised volatility responds quickly and is noisy; a 60-day one is stable and slow. Neither is correct; they answer different questions. What matters is that a comparison between two volatility figures computed on different windows is not a comparison at all, and volatility charts routinely mix them without saying so.
The annualisation factor deserves a moment because it embeds an assumption. Multiplying by the square root of time assumes returns are independent across periods. If returns are serially correlated, which they sometimes are after a shock, the scaling is wrong and the annualised figure misstates the actual variability over a year. It is a good approximation most of the time and it is an approximation.
The distributional facts are worth knowing because they undermine the naive use of the number. Oil returns have fat tails: extreme moves happen far more often than a normal distribution implies. So a volatility figure is a reasonable description of the middle of the distribution and a poor guide to the tails, and any risk calculation that scales volatility into a tail estimate under a normal assumption will understate the tail systematically.
The comparison that makes realised volatility useful is against its own history rather than against an absolute standard. Twenty-five percent annualised means nothing without knowing whether that is high or low for this market. That is what a vol cone does: it plots current realised volatility against the distribution of past realised volatility at the same horizon. See term structure and the cone.