Philosophical "zombies" are imaginary beings that are exactly like us, except that they lack "phenomenal" consciousness -- the kind of consciousness that makes it "like something" for us to see, for instance, a red dot. Whereas our world is phenomenally laden, you might say, the zombie world is phenomenally void.
Call our world (the actual world) "W1" and the zombie world "W2." Call the set of W1 properties in virtue of which W1 is distinct from W2 "Φ1," and the set of W2 properties in virtue of which W2 is distinct from W1 "Φ2." Finally, call the set of all the properties in common between W1 and W2 "Ω."
Now imagine a third world, which we'll call (you guessed it) "W3." W3 is identical to both W1 and W2 in respect of Ω. What distinguishes W3 is that it begins its existential career as a conjunction of Ω and Φ2, but then at some time t undergoes a phase change, whereafter it persists as the conjunction of Ω and Φ1. In other words, W3 is phenomenally void until t, and phenomenally laden thereafter.
Suppose that at t minus 1, George3, an adult denizen of W3, and incidentally a loyal reader of this blog (or rather, its W3 counterpart), looks at the following dot:
George3, at this point still anaqualiac, says aloud: "I am looking at a red dot." (George3, like his otherworldly counterparts, has a habit of talking to himself.)
Still curious (at least apparently, for he is furrowed of brow) about where this whole argument is headed, George3 continues to read on until, at t plus a split second, he reaches the end of this sentence, and again looks at the dot following it:
George3 again notes aloud, "Well, that's the same red dot as before."
And it is the "same" dot. But of course something has changed: George3 is actually conscious of seeing the red dot.
Now, by hypothesis, this seemingly striking change in the way of things makes absolutely no difference to George3's outward behavior: W1, W2 and W3 are all identical in their Ω properties, and those properties are supposed to include (at the very least) all observable behavior. However, it is difficult to imagine that George3, suddenly endowed with phenomenal consciousness, would not react to this fact in some observable way: "OMFG! I can...see!! I can see the dot!! I mean, really see it!!! [Bursts through the front door, laptop in hand.] Look, everybody! Do you see the red dot?! [Now weeping.] Oh...God -- it's so beautiful!" But...no, the zombie hypothesis has George3 going on exactly as before, as if nothing at all had changed.
If the counterintuitiveness of the situation at W3 suggests there is something wrong with the zombie hypothesis, there are at least two lines of response open. The first is simply to concede that W3 is inconceivable, but point out there's no clear reason W3's inconceivability should impede the conceivability of W2.  The second is to argue that the seemingly incongruous behavioral continuity is explained by some isomorphism between W1 and W2 (or W3 before t and W3 after t).
The first line of argument at least seems fishy to me. It is true that the apparent incongruity in W3 is generated by the distinctive "phase change," arguably a feature that makes the distinction between W3 and the others look like an easy call. But I think this move smuggles in the sort of dynamical assumptions that zombie arguments need to bracket out.  The phase change, as I've defined it, doesn't invoke a separate, dynamical, causal element; rather, it simply marks the time of an ontologically brute change in W3's career from the era of phenomenal voidness to the era of phenomenal ladenness. That being so, there doesn't seem to be any principled difference between deploying the zombie concept at W3 and deploying it at W2; W3 is merely a way of looking at the zombie hypothesis in light of further considerations. If so, and if W3 is inconceivable, this counterintuitiveness would seem to provide further reasons for (fatal?) suspicion about the zombie hypothesis.
The second line of response, that of appealing to some relevant isomorphism, might go something like this. A given instance of (pseudo)reference to a quale at W2 supervenes on certain Ω properties, but its outwardly reliable (pseudo)referential success supervenes on some cluster of Φ2 properties, say φ2. This cluster is isomorphic in some way (we will stipulate) to the cluster of "intrinsic" Φ1 properties, say φ1, that subtends the successful reference to the corresponding quale in W1. And it's this isomorphism that grounds the behavioral continuity observed in W3: the ontological substrate φ2 of George3's memory or representation of "red" at all t minus n subsequently, at t, gets "phenomenalized" (i.e., roughly, replaced with φ1), but this φ1-induced phenomenal aspect only "takes the place" that φ2 held, and therefore doesn't "add" anything that should cause George3 to behave differently.
But this tack raises two new concerns. First, it construes "intrinsic" properties, which are supposed to be nonfunctional, in a very functionalistic way: the isomorphism between Φ1 and Φ2 is cast as the "normalizing" factor in George's behavior.
Second, it is in tension with a view about phenomenal judgment, advanced by David Chalmers, according to which our knowledge claims regarding qualia are justified all and only by the "intimate epistemic relation" we have to our qualia experience. Here, however, the justification for George3's post-t judgment about red can't be justified by his acquaintance with it, because by hypothesis he'd never before been acquainted with it. Yet (for all that) George3's judgment seems perfectly justified. If that's right, then zombie advocates owe us some other account of justification for phenomenal judgments.
1. I am ambivalent about whether zombies are ideally conceivable, but have argued elsewhere that to whatever extent they are, their conceivability only undermines the force of the argument from the possibility of zombies to the potential explanatory efficacy of "nonphysical" or "intrinsic" properties.
2. See Peter Bokulich's "Putting Zombies to Rest: The Role of Dynamics in Reduction" (pdf).