What impedance?
The usual method is a LISN, or CDN, or bias tee: a series inductor isolates the DC supply from the AC signal and EUT, and a series coupling capacitor isolates EUT voltage from the signal.
If the EUT should see an impedance of 50Ω, then this is adequate. If it's a much lower impedance -- as might be the case for a related kind of test, like power supply upset, or power interruption attack (see ChipWhisperer and such), then consider merely using a resistor divider, and much more signal power, to get the desired supply impedance and signal input. Upside to resistor divider is, you get attenuation for free, meaning your amplifier doesn't see a nasty SWR, regardless of what the EUT impedance is. Downside is, you can end up needing 10 or 100W of RF, very easily...
At some point, attenuation gets ridiculous, and you should use a transformer instead. Or a combination of attenuators and transformers, as suits the system, for frequency response, SWR, available power, required stimulus, and so on. Just a compromise between these, however it works out best.
An astute observer may note: "Ah, but you just suggested using an inductor, coupled to a transformer -- can't that just be reduced to a transformer?"
Indeed it can. The coupled configuration is called "parafeed" (in audiophile circles, at least). One advantage is using all off the shelf components: no worry about DC in the transformer. If you're winding it yourself, yes, you can gap the core to get the required DC handling, or add the primary winding to the inductor just as well; but this is undesirable for RF response, because you need a much longer winding to get the same impedance. Winding length corresponds directly to the HF cutoff point.
For modest bandwidths (say, 2-3 decades), this will be fine. Example (again from audiophilery), vacuum tube output transformers (single ended output type) typically satisfy the 20-20kHz frequency range, while handling an equivalent amount of DC as signal (i.e., typical of a class A amplifier). Using a much smaller winding on gapped ferrite, say, you can easily recreate this experience, but at, say, 10kHz to 10MHz.
Note that exotic materials don't have any benefit here: nanocrystalline cores, or high-mu ferrite, are just thrown away because of the air gap. Powdered iron is quite reasonable, but don't go too low on permeability. Probably 30-100 mu is best, so as to keep the winding length on the short side, without requiring an excessively large core.
As for completely different methods entirely: you may well find it's better to build, what is essentially an audio amplifier, or regulated power supply, that includes your desired bandwidth. 50Hz is too low for practical ferrite inductors, and 200kHz is too high for practical laminated-iron inductors. (So obviously, stack both to make a wideband bias tee -- this requires some finagling, however.) The nearly 4-decade range is quite reasonable for nanocrystalline-core transformers, but you may find it's easier to avoid going that route, and solving it with transistors instead, even despite the high current.
Tim