The honest answer depends much more on distance and duty cycle than on the rectifier alone, and that is where most demos are misleading.
With a dedicated transmitter a few metres away, I have seen tens to a few hundred microwatts of usable DC power after rectification, enough for a sensor node that wakes up, takes a measurement, sends a short packet, and sleeps again. Move that same node to an ambient environment, a few metres from a Wi-Fi router or much further from a base station, and the available power usually falls to a few microwatts or less. At that level the node is not continuously powered; it is charging a capacitor or small supercapacitor for seconds or minutes, then spending the energy in a millisecond burst.
Two effects set the floor. First, rectifier efficiency collapses at low input power because of the diode threshold and the matching network loss. A rectifier that is 50 percent efficient at 0 dBm may be below 10 percent at minus 20 dBm, so quoting an open-circuit voltage tells you almost nothing. Second, a broadband match collects energy from more sources but is less efficient at any one frequency, while a narrowband resonant match is efficient only when the antenna, frequency and position are all right.
My practical test is to report the loaded DC power against distance for a known transmitter, together with the load current and the wake-up duty cycle. If the average load is 10 microwatts and the harvester delivers 5 microwatts at the deployment distance, the node will slowly lose charge no matter how good the peak number looks. If the harvester delivers 20 microwatts at that distance, the same node can run indefinitely with a sensible sleep schedule. That power-versus-distance curve, measured into a real load, is the number I trust.