In the process, some magnetic field goes into the metal, inducing a current (which is how the field reflects off it, in the first place) and thus causing it to heat up.
As seen by the circuit, the effect is a lower work coil inductance and modest Q factor, whereas steel has similar inductance and low Q factor.
Generally speaking, for similar setup, magnetic steel gives a lower Q than stainless, than aluminum, than copper.
In any case, it's just an impedance matching problem. Which is why I've presumed "they've designed it to be compatible with aluminum and copper-bottom pans". Simply check if it's rated for that or not, and use whatever metal is appropriate.
Like I said (or hinted at), induction is perfectly reasonable on PCB material as well. In fact I've used this to tin large areas of copper clad before, just set up an open-face ferrite core on the breadboard, with a half-bridge driver running from a function generator. Direct drive or series resonant is very reasonable, and as it happens, fixed frequency is pretty easy to control.
Typical Q for very conductive metals, with close coupling, is 10-20. Meanwhile the inductance drops in proportion to the amount of field being shielded (which may be quite considerable for "close coupling", anywhere from -10 to -80% say?). Best Q for something like copper foil (like PCBs) is around 6 or 7, from what I've measured. This is at frequencies where skin depth is about equal to foil thickness, so the shielding is relatively poor and the foil looks somewhat more resistive than reflective. For similar conditions on magnetic steel, it can be 2-5, quite a lot of resistance indeed -- direct drive (no resonant cap needed) is even feasible.
Typically the kind of induction heaters that heat up or even melt aluminium or copper or gold or whatever use a air core coil (allows for stronger fields) and place the work piece right in the middle of the coil rather than on top (solves those magnetic coupling issues)
Mainly it's that the frequency is higher (which is neither here nor there, really), the high temperatures, and requirement for insulation, preclude use of most any core material, which would get very hot from core loss anyway (requiring water cooling just as the coil does). Field strength is also secondary; it's simply about how much power you can put into the coil, and its efficiency. Though the pole pieces used with cooktops do help greatly to improve the field strength at the surface; this helps reduce Q, and also shields the circuitry underneath, an important feature.
That said, there are some cored applications in industry. Channel induction furnaces use a loop of molten metal, essentially as a shorted turn, around a laminated steel core, and operate at mains frequency (plain old phase control is feasible, though hopefully they're doing something a bit smoother for sake of power factor). These are typically in the low MW power range.
Overall equivalent figures like L and Q are more important than particulars like field strength and skin depth. Things are a bit more particular if you're doing stuff like case hardening (the skin depth and thus frequency does matter, albeit not very sharply because skin depth depends on sqrt(f)), or melt stirring (lower frequencies create more MHD and convective action).
(For those unaware, I've been designing induction heaters for many years; this comes from direct experience.)
Tim