The obvious equilibrium is that more researchers can download papers from the internet, and thus we expect more papers to be read by a greater number of people. If lay people enter the calculus, this is almost certainly true. But what about researchers? I am not convinced that more reading (of each paper) goes on, or that it should go on.
Most people, including researchers, cannot easily figure out if the main result of a research paper is correct. That is true all the more as time passes, because the mistakes become less and less transparent. But they can figure out who can figure out if the paper is right, and sample that opinion. The internet aids this process greatly. For instance, it is easier for me to find out what Bob Hall (one of the great paper analysts/commentators of all time) thought of a macro paper, if only by using email. If I can find out whether or not the paper is true, often I don’t have to read that paper, though I may go through some parts of it. The internet also gives me access to better summaries of the paper, if only in parts of other papers.
In this sense, researchers may rely on a fairly thin substructure of evaluation, though one of increasing accuracy. As science progresses, perhaps scientists do/should spend more time honing their research specializations, and less time reading papers they are not expert evaluators for. They do/should spend more time reading the papers where they are the expert evaluators, but that may mean reading fewer papers overall.
Viewed as a productivity problem, perhaps your read is competing against “further spread of the read and evaluation from the best expert” and is losing. Efficient criticism is also sometimes winner take all.
I am indebted to Patrick Collison for a conversation on this topic, though of course he is not liable for any of this. Neither he nor I have read a paper on such matters, however. Thank goodness.