Methods and apparatus for audio watermarking are disclosed. Example methods disclosed herein include obtaining a watermarked noise signal having energy in a first frequency band, the watermarked noise signal being associated with a first operating state of a device. Disclosed example methods also include adjusting the watermarked noise signal based on an ambient audio level sensed by an audio sensor. Disclosed example methods further include emitting the adjusted watermarked noise signal via a speaker in response to determining that the device is in the first operating state.