How big is science's fake-paper problem?

In support of the above comment and from the linked Nature article:

    The analysis estimates that 1.5–2% of all scientific papers published in 2022 closely resemble paper-mill works. Among biology and medicine papers, the rate rises to 3%.

The problem mainly lives in the "paper churn" "publish or perish" "look at us, we're relavant" domain, those papers that to be brutally honest are rarely read and have little impact.

Papers that start to be taken seriously and have any kind of impact don't last long if they've been faked, sooner or later some one will declare a problem with getting the same results or try to gain attention by exploring a different angle that reveals issues with the (assumed) foundational paper.