The 95% confidence interval says that "if we repeat this experiment again, the estimated mean will have a 95% chance of falling within this range".

This relies upon the Central Limit Theorem, which says that sample means are normally distributed even if the variable is not normally distributed.