Optical pulse sensing has arrived in earbuds, including models aimed at sleep tracking that report heart rate and heart-rate variability through the night. The ear is a reasonable place to measure it, and a poor place to measure some of the things people will assume are being measured.
The technique is photoplethysmography: illuminate tissue, read back the light that returns, and recover the small periodic change caused by blood volume with each beat. The ear canal helps here. It is well perfused, it is shielded from ambient light, and it moves much less than a wrist does during sleep, which removes most of the motion artefact that wrist devices spend their signal processing fighting.
What the signal supports reliably is beat-to-beat timing: heart rate, and from the interval series, variability. That is enough for sleep-stage estimation and for trend tracking across weeks. What it does not support is anything requiring absolute accuracy of the optical amplitude. Blood pressure from pulse shape alone remains unvalidated outside controlled studies. Oxygen saturation needs two wavelengths, careful calibration and a known optical path, and an in-ear measurement of it should be treated as indicative at best.
If you are specifying such a product, three things decide whether it works: contact stability across ear shapes, ambient-light rejection at the photodiode, and an honest confidence output. A sensor that reports a number regardless of signal quality will produce a plausible heart rate from a bud sitting on a desk, and users will believe it.
Battery is the other constraint worth stating early. An optical sensor runs a light source continuously, and a night of measurement is a far heavier duty cycle than a few hours of listening, so the sensing budget competes directly with playback time.