The slowest comparison sort, and the clearest one to reason about.
Bubble sort walks the array from left to right and swaps any two neighbouring values that are out of order. Each element that is too large therefore slides rightward one place per pass, and after a full pass the largest remaining value has reached its final position at the right end.
The name describes what you would see if you watched the values: smaller ones rise through the list like bubbles. Nothing about the algorithm depends on that imagery — what matters is the invariant that after pass k, the k largest values are in their final positions, so the unsorted region shrinks by exactly one per pass.
The naive version always runs the full n-1 passes, which is quadratic even on sorted input. Adding a flag that records whether a pass performed any swap makes the common case cheap: if a whole pass makes no swap, the array is already ordered and the sort can stop immediately.
That single flag is the difference between O(n^2) and O(n) on already-sorted or nearly-sorted data, which is precisely the case that makes bubble sort interesting in practice as well as on paper.
Almost nothing, in terms of speed. It appears in teaching because the invariant is obvious and the visualisation is immediate — which is why this page ships an interactive one.
It is worth knowing as a contrast case. It is stable, it sorts in place, and it is the only sort here whose correctness argument needs no auxiliary structure, so it is a useful baseline against which quicksort's partitioning and heapsort's heap property are compared.