Robert Columbia's framing of the question:
It seems that you are not asking about the ethics of downloading articles from questionable sites, but only the ethics afterwards - when the downloading is already a fait accompli, and the real question is whether citing said document counts as an additional unethical act above and beyond what you did when you downloaded it (e.g. "citing a document obtained through an unauthorized route").
For discussion purposes, I'm going to go contrary to the accepted answer and say yes.
First, if you're using ideas that are not your own, you definitely have to cite the source. Failure to do that is plagiarism. With respect to avoiding plagiarism, it doesn't matter where you got the ideas; credit is due.
However, inaccurately citing where you got the source makes it harder for others to trace that back to the source.
In some cases, the content posted to a pirate website looks like or maybe is labeled as a copy of the content from the original source, but in fact has different content.
I have wasted many hours trying to figure out why an author cites X for assertion Y, when in some cases they were actually citing a hidden X' which differed from X on, among other things, assertion Y. I would have much appreciated if the author had included a citation like:
J. Ixsom, P. Smiflich, and D. Seuss: "Methods for safely pipetting oxyflogated dexlahydrates at STP." 23rd Annual Conference on Unusual Chemistry (CUC '17) London. Retrieved Jan. 18, 2019 from https://authorsite.org/pdf/1234.56789.
Then I have all the usual citation information about who published that and where, and in most cases I can go to that source and get it from them. In this case, the publisher gets an extra sale from the author having cited that work, regardless of how the author obtained it. However, if I can't find support for what's being asserted, I can then trace through to what the author actually looked at. If there's a difference, I can more quickly debug. I might see how the author was led astray and be able to resolve the issue another way. Without that link to what the author is actually citing, though, it's very hard to determine the basis for that assertion, and requiring that extra work is a negative impact caused by a difference between the source the author claimed and the source the author actually used.
Also, note that content differences are not all malicious, and the "authoritative" record is not always better from a content perspective. Sometimes, authors posting a file on their own website, with the publisher's permission, will update those files (e.g. correcting a mistake in a formula) and then the formula being used in the citing article looks different than the one it's being cited for. (To authors: If you do this, please explicitly call it out as a correction.)
Sometimes, the differences are honeypots intentionally inserted by publishers designed to get the message out that only official sites can be relied upon for accurate information.