Five salaries, in thousands: 45, 48, 50, 52 and 300. The mean is 99. The median is 50.
Both are correct. Both are called the average in ordinary speech. Four of the five people earn less than half the mean, so if you reported "the average salary here is 99" you would have said something true and left everyone with the wrong picture.
What each one is actually measuring
The mean is the balance point of the data: add everything up, divide by the count. Every value pulls on it, and values far from the centre pull hardest. The median is the middle value once the data is sorted, and it only cares about position, not distance. Move that 300 to 3,000 and the mean leaps while the median does not move at all.
That difference is the useful signal. Run the same five numbers with the outlier replaced by 55, and both the mean and the median come out at 50. When the two agree, the data is roughly symmetric. When they diverge, something is stretching one tail.
Reading the gap between them
The direction of the gap tells you which way the tail runs. A mean above the median means a long right tail, which is the usual shape for income, house prices, page views and repair bills: a floor near zero, no ceiling, and a handful of very large values. A mean below the median means the long tail runs left, which is rarer but shows up in things like exam scores near a ceiling.
The descriptive statistics calculator prints the mean and median together for exactly this reason, and the skewness calculator puts a number on the asymmetry. For that salary set it returns 2.233, firmly right-skewed.
Deciding whether the tail is real
A value far from the rest is not automatically an error. It might be the most important observation you have. The question is whether it belongs to the same population as everything else, and that is a judgement about where the data came from, not one the arithmetic can settle.
What the arithmetic can do is flag which points are unusual enough to warrant asking. The outlier fences calculator applies the standard interquartile rule; on those five salaries it puts the fences at 42 and 58 and flags the 300.
A practical rule
Report the median when the distribution has a long tail and you want a figure that describes a typical case. Report the mean when totals matter, because the mean multiplied by the count gives back the sum, and the median does not. A payroll budget needs the mean. A description of what people earn needs the median.
Where the two disagree substantially, the honest answer is to give both. One number was never going to describe that data on its own.