Riordan et al. was retracted in November 2021. By June 2026 the paper, a stem-cell clinic study, had collected 72 citations on Google Scholar, and 27 of them arrived after the retraction notice went up. Of the citations logged since 2023, 15 do not mention that the paper was ever pulled.
Those numbers come from a count that Jeremy Snyder of Simon Fraser University and Leigh Turner of UC Irvine published in September in Stem Cells Translational Medicine. They call the phenomenon “zombie papers”: studies formally withdrawn from the scientific record that keep walking through it anyway, cited, footnoted, and built upon as if the withdrawal never happened. A retraction is supposed to function as a death certificate for a finding. In practice it is a label almost nobody downstream ever reads.
Snyder and Turner did not reach for Riordan et al. at random. It was a clinic publication that the journal itself retracted, and both authors have spent years documenting how the unproven stem-cell industry uses the published literature as sales material. They log a second retracted study alongside it, Patil et al., pulled in February 2019 for image duplication, which has collected 13 citations since 2020. None acknowledge the retraction.
The point of the count is not two papers. It is the ratio. The publicly searchable Retraction Watch Database now lists more than 67,000 retractions, and in stem-cell publishing alone Snyder and Turner report the retraction rate reached 0.84 percent in 2023, 117 of 13,856 papers. Each of those becomes a candidate for an afterlife, and the afterlife is well measured: an analysis of biomedical literature in Quantitative Science Studies found that of 13,252 citation contexts that referenced a retracted paper after its retraction, only 722, or 5.4 percent, acknowledged that the paper had been retracted. The withdrawal happens. The knowledge of the withdrawal does not travel with it.
Follow the incentive, because it explains why nobody is in a hurry to fix this. A citation in a real, indexed journal is an asset, and it does not stop being an asset when the paper is pulled. For a clinic selling an unlicensed injection, a footnote to a peer-reviewed study is a marketing instrument, and a reader who checks that footnote lands on a paper that, in most databases, still looks like a paper. Turner and Snyder have tracked exactly this marketplace, where peer-reviewed citations become sales collateral: their earlier work catalogued crowdfunding campaigns for unproven stem-cell interventions that cited published efficacy claims while burying the risks. A retracted paper that keeps its citations intact keeps its usefulness intact.
No cited source ties Riordan et al. to a specific clinic closing a specific sale, and Snyder and Turner do not claim one. What the record establishes is the citation counts, the acknowledgment rate, and two clinic-adjacent papers that kept drawing references for years after they were withdrawn: a mechanism and an incentive pointing the same direction, with no counter-force strong enough to reverse it.
The counter-force is missing because responsibility for a retracted paper is distributed until it belongs to no one. The cheapest fix sits with the databases and search tools, which could flag the status at the point a reader finds the paper. The publishers control the article page and could carry a “retracted” label into the title and watermark the PDF, and could preserve that label when a journal changes hands and its back catalogue is transferred to a new platform. Reference managers such as Zotero could warn an author before a bibliography is submitted; journals could check the retraction status of cited works during peer review; the generative AI systems now resurfacing old papers could refuse retracted sources. None of this is a research problem. Each fix is a decision about who owns the cleanup, and the parties that benefit most from the citation staying intact are the ones selling what the paper appeared to support.
Formal retraction is only the visible half of the problem. The same Retraction Watch roundup that surfaced the zombie-paper analysis also flagged a ghostwritten pain-management paper, described in one journal account as a “twisted phantasmagory” of academic writing, and a hijacked-journal checker that has now logged more than 450 entries. A ghostwritten paper never earns a retraction to ignore. It entered the record dishonest and stays there, citable, doing the same work a zombie paper does.
The dataset that could flag all 67,000 exists, and it is public. Whether a reader ever sees the flag is decided somewhere else: by the publisher that controls the article page, the indexer that controls the search result, and the reference manager that controls the citation as it is pasted into the next manuscript. Snyder and Turner named each of those parties and the fix within each one’s reach. As of the September issue, the count on Riordan et al. was still climbing, and the cleanup still belonged to no one.
Sources
- Stem Cells Translational Medicine – Snyder & Turner, “Why retracted articles can have an influential afterlife, and what to do about it” (Sept 2026)
- Quantitative Science Studies (MIT Press) – Continued use of retracted papers: citation contexts and awareness of retractions in biomedicine
- Retraction Watch – Weekend reads: the afterlife of retracted articles; ghostwriting; database counts (Sept 12, 2026)
- The Niche (ipscell.com) – Questions remain after Stem Cells Translational Medicine retracts clinic publication (Dec 2021)
- STAT – Crowdfunding for unproven stem cell procedures wastes money and spreads misinformation