Especially in the earlier days of Ask Patents many posts used "prior art" as synonymous with "killer prior art that would completely show a patent document to not be new". Technically "A is prior art to B" just means that A is allowed to be used against B. To qualify under this correct, and low, hurdle A needs to meet a timing criteria, an "available to the public" or "published" criteria, and might need to meet a technology field criteria. The arcane rules that split the hairs of what is in and what is out are set by statue and by precedent or "judge made" law.
Anything published anywhere anytime in any language can be prior art. Something written down and put in a drawer is not published and therefore not prior art to anything. The Gutenberg bible can be prior art your inventors notebook kept in your garage can't.
Crudely, A can't be prior art to B unless A came before B. However, if the inventor of B says A is prior art on the record, then it is "admitted prior art" even if it would not otherwise actually qualify. In terms of technology field, prior art must be from the same or an analogous art. Or be from any field but address the same problem.
Now what is "good" prior art? Patents and patent applications typically have multiple claims. You can think of each claim as a mini-patent in that they are individually rejected or allowed. And individually found valid or invalid, and likewise individually infringed or not infringed. Among the several requirements for patentalbity, the territory defined within the bounds of a claim must be new and non-obvious. Prior art is used to show that a claim is not new or it is obvious.
It is not too hard to be new. A claim says it covers a device that includes a widget connected to a large thingy on a base with a flat bottom. It is shown to be not new if anything that minimally fits that description has been "taught" in a prior art qualifying document. I say minimally because the prior art might have also described the device has having a spherical widget or specified that the device also has an antenna. It still does have a widget connected to a larger thingy on a base with a flat bottom. If the bottom weren't flat it would not show the claim as old.
If one single document describes a device with all the elements of the claim and the device has all of the specified interconnects or other relationships between the elements in the claim then we say that document "anticipates" the claim. Meaning the claimed territory has at least some region that is not new. No allowance for that claim.
If the examiner can't find a single prior art document that describes everything the claim claims they look for prior art that contains most of what the claim encompasses. The claim is an X connected to a Y with a Z on top. The examiner can only find something with an X connected to a Y with a W on top. It does not prove the claimed invention old. So the examiner finds something that says in that field W and Z are known to have similar properties or they find something else with no X but it has a Y with a Z on top. They argue that it would be obvious (they need actual logic here) to combine the two to get the claimed invention.
That does not make the claimed invention not new. It makes it obvious. Unlike some posters, the examiner can't wave their hands and say it seems obvious to them. They need two or more documents that, between them, have all of the elements of the claim and then need some logic as to why the problem at hand or "market forces" would drive someone of "ordinary skill in the art" to put them together to achieve the claimed invention. Of course this is inherently subjective.
I have been talking about individual claims. Prior art that shows claim 1 to be old or obvious may not show claim 10 as old or obvious.