No single organization checks research integrity end to end. Different groups check different things at different points in a paper's life. Before publication, participating journals can run a submission through platforms such as the STM Integrity Hub and Clear Skies' Papermill Alarm, which flag it for closer editorial review. After publication, the Crossref Retraction Watch dataset and PubPeer's disclosed comment threads let anyone check what has already surfaced about a paper. Lucent runs a narrower, reader-facing check, Crossref's own retraction metadata, on every paper it covers. Below: what each one actually does, sourced, and what none of them check.
Screening, where it happens, mostly happens behind the editorial curtain. A manuscript can pass through an image check, a paper-mill pattern check, and a duplicate-submission check before a journal ever accepts it, and a reader who later opens the finished, published paper has no way to see any of that. There is no badge on a PDF that says "screened by the STM Integrity Hub" or "checked with Proofig." The checks exist. They are simply invisible from the reader's side of the page. This page names who does what, so that gap is at least legible.
The STM Integrity Hub is a shared, cloud-based screening environment that the STM Association built for publishers to check submitted manuscripts for research-integrity concerns before acceptance. According to an STM Association report dated September 12, 2025, more than 35 publishers use the Hub to screen over 125,000 papers a month, and the Hub flags around 1,000 of those as resembling known paper-mill output. It works as an integration layer: publishers opt in to individual tools inside it rather than one fixed checklist.
One of those tools is Clear Skies' Papermill Alarm, first launched as a public tool in 2022 and incorporated into the STM Integrity Hub under an announcement dated September 1, 2023. It compares a submitted manuscript's text and structure against patterns found in publications already linked to paper mills, then returns a traffic-light rating: red for high similarity, orange for moderate similarity, green for none. The rating is meant to help an editor decide where to look closer, not to stand in for that review.
Image screening is a separate check again. Proofig scans a manuscript's submitted figures for duplication, rotation, splicing, and other alterations by comparing them against a large database of previously published images. Science's editorial family adopted it across all six of its journals, according to Proofig's own announcement dated January 4, 2024. None of these three tools are used everywhere: adoption is opt-in and journal by journal, so a paper's presence in a given journal is not proof it passed through any specific one of them.
Once a paper is out, the record of what happens to it lives in different places. The Crossref Retraction Watch dataset is a public file that Retraction Watch keeps building and Crossref hosts and rebuilds most working days, recording each retraction as a dated row. Retraction Watch also runs its own searchable database on the same underlying data, built for browsing a journal, an institution, or a topic rather than checking one DOI at a time. We cover exactly how large that dataset is and how fast it is growing on how many papers get retracted.
PubPeer is different in kind again: a site where readers and researchers can leave dated comments on a published paper. Its own posting rules restrict comments to publicly verifiable facts. A thread existing on PubPeer is a record that a fact-based comment was posted and disclosed, publicly, at a given time. It is not a finding, not a rating, and not a verdict on the paper, and Lucent treats it exactly that way: a disclosed thread to read for yourself, never a conclusion to repeat.
Lucent is not a competitor to any organization named above, and does not try to replicate what they do. Every paper we cover gets one narrower, reader-facing check: we confirm the DOI resolves, live, and we read Crossref's metadata for a retraction-related field, then we show readers exactly what was checked in plain words next to the paper, not a score. The full method is written out at how Lucent checks every paper before you read it, and anyone can run the same lookup themselves for free at Lucent's DOI checker. We sit downstream of the same public infrastructure named above: Crossref's data, made visible to a reader who will never open the STM Integrity Hub or a publisher's editorial system.
Paste it and get a plain-words answer, sourced to Crossref, in seconds.
Run the check at /checkEvery tool here has a stated boundary, and stating it is part of using the tool honestly. The STM Integrity Hub and Papermill Alarm work from patterns in a submitted manuscript, so a genuinely novel case that resembles nothing seen before is the hardest kind to flag by pattern-matching, by construction. Proofig checks the images actually submitted, not whether the experiment behind them was conducted as described. The Crossref Retraction Watch dataset and Retraction Watch's own database can only record a retraction after a journal has issued one, so a real, unresolved problem with a paper that has not yet been retracted will not appear there, sometimes for a long stretch of time: see how to check if a paper was retracted for what that lag means in practice. PubPeer only surfaces what someone has chosen to post. And Lucent's own check confirms two narrow facts, a resolving DOI and the absence of a retraction-related Crossref field, at the moment we looked; it says nothing about a paper's methods, its images, or its conclusions. An absent flag anywhere on this page is not the same thing as a clean paper. It means that particular check did not find that particular problem, today.
None of these steps requires an account, and none of them is a substitute for reading the paper's own methods section. They are ways to see what has already been publicly checked or disclosed, nothing more.
Every two days: brand-new papers in plain language, each checked for a live DOI and a Crossref retraction flag before you ever see them. Three free every issue.
How Lucent checks every paper · Check a DOI · How to check if a paper was retracted · How many papers get retracted · Latest issue · Privacy