I'd call an ad-hoc antenna that works on multiple bands with an ATU a "Kinda-Random Antenna" (KRA). (Apologies to linguistic purists.)
A simplified definition might be: A. long enough to work within the maximum limits of the ATU's L-network on the lowest band used B. presents a reasonably low impedance on all bands used (e.g., doesn't look like an end-fed half-wave) ATUs have limited monotonicity and granularity, as well as stray impedances, so in practice there is a third criteria: C. tunable on each band used despite specific L-network idiosyncrasies This third criteria is the hardest one to predict for a given ATU design, as the idiosyncrasies vary with PCB layout and actual component values. They may only impact the highest bands, or for a particular antenna, the bands on which Q is the highest. For our ATU designs, we try to minimize strays and keep the network monotonic by using tightly toleranced capacitors and toroidal inductors. While a wide range of wire lengths will meet the requirements of a "KRA" in the field, we've found from experience that something in the 25'-28'(7.5-8.5 m) range works on all bands from 40 meters up, and roughly twice (15-17 m) this for 80 meters up.
Since it's impossible to predict the effect of ground losses, obstructions, deployed wire angles, etc., you may occasionally need to add or remove wire to obtain resonance on all bands used. 73, Wayne N6KR