Forensic analysis, and with respect to ancient documents, particularly palaeography, the scientific examination of ancient documents - reveals that none of manuscripts we have are old enough to be autographs, based on our knowledge of when the actual events occurred.
See: Rylands Library Papyrus P52.
What more direct evidence do you think we could have, being almost 2000 years after the actual events? All copies signed with names of the scribes who copied the material? But that is not the case. That someone attested to destroying all the originals?
As with all of our studies of ancient texts, artifacts, ruins, mounds, etc, we determine their nature by comparing our forensic analysis of the artifact itself with the archaeological, historical and anthropological record and context in which they were found and refer to. That method represents our "state of the art", and our advances in this field are simply advances in our ability to do more accurate forensic analysis and contextual juxtaposition.
That is really the best we can ever hope to do (at least until we manage to perfect safe time travel...) when dealing with material from the remote past, barring exceptional cases where we are lucky enough to happen upon something that yields direct first hand information. In the case of NT, I think it is safe to say that we have no such first hand information, or it would be headline material.