Some researchers based in India suspected there might be an additional reason retracted papers are still being cited: Sci-Hub. Sci-Hub works a bit like a combination of cache and aggregator for published materials. Whenever it gets a request for a paper that’s not already in its database, it uses leaked login credentials to go to the website of whatever journal published the paper and obtain a copy. If it already has a copy, however, it will simply serve that up instead. This leaves open the possibility that it will have obtained a copy of a paper prior to its retraction and continue to distribute that copy after the paper has been retracted.
To check this, the researchers obtained a list of nearly 17,000 retracted papers and searched for them on Sci-Hub. They then visually examined the documents that were returned. They found that 85 percent of them contained no indication that the paper had been retracted. “The availability of [unlabeled retracted articles] in the field of health sciences is particularly high,” they note, “which indicates a significant risk of their unintended use and further citation in future research.
Staying up to date
While corrections are less severe than retractions, they’re likely to suffer a similar problem. And corrections will often involve the technical details of a paper—the experimental approaches or raw data that will be critical for anyone wanting to replicate or extend previously published results. So, if anything, their impact will be more significant.
It’s worth noting that this was already recognized as a problem even in the absence of Sci-Hub, and a system called Crossmark was developed to make it easy to find the most up-to-date version of a paper, including any corrections or retraction notices. Of course, not all publishers use Crossmark, and making it easy to use doesn’t guarantee researchers will find the time to go through what might be a massive library of reference papers and check each one before publishing.
But, given that Sci-Hub is an automated system, it doesn’t suffer from the lack of free time or motivation of an actual living researcher. In theory, there’s nothing to prevent it from being updated to incorporate a check of whether its cache contains the most recent version of a paper whenever it’s requested.
Accountability in Research, 2025. DOI: 10.1080/08989621.2024.2446558 (About DOIs).
