What Frequency is used?
Many codecs sample at 44.1 KHz or lower (especially web based ads). Given Nyquist's (or Nyquist-Shannon as it's now know) theorem (max theoretically reproducible signal is half the sample rate), how does the ultrasonic signal get reproduced?
Also, many speakers do quite poorly over 18KHz, and many microphones aren't good with high frequencies, either, so even if an ultrasonic is (barely) present, "hard of hearing" (in higher ranges) microphones might not hear them.
Seems like this ad tracking technology has some serious roadblocks in its way, just based on the current technology. Might do better with some sort of infrasound, or audible but unrecognizable as a signal sound (audible steganography) for tagging ads.
Or - even better, audible watermarking of sound sources with date/time/channel ID stamps -- instead of tagging the ads themselves, tag the medium. After the spyware phones home, the big data analytics engines can do the mass lookups to correlate date/time/channel stamps with what content was playing. This way, individual advertisers don't have to "play" to be useful. Some sort of "ignorable" (by humans) noise would carry the information.