Binary search IS efficient (O(log n)) even for large datasets - its power grows with size. The real limitations are: A) it requires a sorted array, B) maintaining sorted order is costly with frequent insertions/deletions (requires resorting or complex data structures), and C) it needs direct middle element access (random access, not suitable for linked lists). Option D is NOT a limitation.