There’s a way to spot data fakery. All journals should be using it