Kevin Hoover writes: "Whenever, we construct the curves that feed into models with well defined equilibria, we do it conjecturally.  The demand curve is, for example, the locus of quantities that people would buy, ceteris paribus, as the price changes ... The case with the aggregate-demand curve is no different.".
 
Every p-q combination along a demand curve has (a) a respectable theoretical underpining (utility maximisation) and (b) corresponds to an observable or at least plausible set of relative prices.  It is worthwhile asking what the demand for good y would be if the price of good y doubled (or was cut in half), ceteris paribus.  We are not generally interested in implausible values of p.   
 
But what is the theoretical underpinning for the thought experiment of keeping M constant and doubling (or cutting in half) P?  Are these potentially observable or feasible outcomes?  Also, if P was arbitrarily cut in half or doubled (M held constant), what would happen to AS?  
 
There is a section of an AD curve that (a) may have respectable theoretical underpinnings (there is some relationship between M and P - its precise form can be argued about) but (b) also has limits (determined by feasible variations in M/P).  Outside that section (the zones of implausible variations in M/P) represents a science fiction conjecture.
 
Robert Leeson