Real studies, honestly described
Peer review is not proof. Paper mills sell fabricated studies that pass as research, retracted papers go on being cited for years, and AI tools invent references that look perfectly real. Pulse Brain holds 38,000+ studies, and every one is checked so that the evidence you build on is real, current and correctly described.
The film, in under three minutes: the scale of paper-mill fraud, the difference between a faked study and one an AI invented, and the five checks every Pulse Brain study passes. Captions are on; the page below says the same in full.
The problem
Paper mills are businesses that produce manuscripts to order and sell authorship on them. Their papers can carry a real DOI, sit in a real journal and pass peer review, sometimes because the review itself was faked or a guest editor was compromised. Publishers have retracted tens of thousands of them: one publisher alone has retracted 11,098 papers for paper-mill activity or compromised peer review.
Paid citations make a weak or fake paper look established. A citation count can be bought, so on its own it is not evidence that a study is sound.
Invented and inaccurate citations are the quieter problem: a reference to a paper that does not exist, a real paper given the wrong authors or year, or a DOI that points to a different study. Reading lists, reports and AI tools all produce them.
What we check
Is it real?
Every study's DOI is resolved at Crossref and its title, authors and year compared with ours. A study without a DOI is searched for at Crossref and OpenAlex. Anything we cannot find in two independent sources is checked by hand, and a study we cannot verify is not shown. This runs every two hours across the whole catalogue.
Is it still standing?
Every study is re-checked against the Retraction Watch database, which Crossref publishes openly. A retracted study is taken out of search, reports and Pulse answers at once. Its page stays, with the retraction notice stated first, so anyone who meets it elsewhere can see it was withdrawn.
Was it published where paper mills were active?
For more than 150 journals hit by paper-mill retractions, we know how many of the papers each published in each year were later retracted for paper-mill activity or compromised peer review. A study from a journal-year where that rate is high is flagged. It is a year, not a journal: many of these journals were sound for years before mills found them.
Does it read like paper-mill text?
Mills disguise copied writing by swapping words for odd synonyms, so
soil fertility
becomessoil ripeness
andartificial intelligence
becomescounterfeit consciousness
. Researchers call thesetortured phrases
. We screen every title and abstract against the list the Problematic Paper Screener maintains.Is the support it gets earned?
A study with a paper-mill signal earns nothing for its citation count, because citations can be bought. And in a Pulse Brain report, a flagged study is never the only support for a claim: the claim stays only where another, unflagged study supports it, and the report says which study carries the flag.
What happens when a check fires
| Signal | What you see | What we do |
|---|---|---|
| Retracted | The study's page leads with the retraction and links the notice. | Out of search, reports and Pulse answers. A new retracted paper is refused at the door. |
| Two or more tortured phrases | The paper never reaches the catalogue. | Refused when it is proposed. |
| High-risk journal-year, one tortured phrase, or an expression of concern | An integrity note on the study's page saying what the signal is and where it comes from. | Listed for an editor to read in full, then kept with a recorded reason, or archived. Never the only support for a claim in a report. |
| Elevated journal-year | An integrity note on the study's page. | Used, and labelled as such in reports. |
An archived study is moved, never deleted: the full record is kept with the reason, and can be restored if the paper is cleared.
What we don't do
We don't judge a study by where its authors work. Paper mills operate in many countries and sell authorship to researchers everywhere. Some countries have higher retraction rates than others, but a country says almost nothing about any one paper, and excluding it would wrong the honest majority of its researchers. Every signal we use is about the paper and where it was published.
We don't treat a signal as a verdict. A flagged journal-year means the paper deserves a careful read, not that it is fake. Only a retraction, or text that is plainly mill-written, keeps a paper out without a person reading it first.
The limits, stated plainly
No screen catches every fake paper. Mills also publish in large, reputable journals, where only a check on the paper itself can find them, and they change their wording once a phrase becomes known. Retractions often come years after publication, so the most recent years are undercounted. We check abstracts, not full texts, and not every study has an abstract available. That is why the checks run every day, why the journal and phrase lists are refreshed from their sources, and why a report never rests a claim on one flagged study.
If you find a study in Pulse Brain you think is unsound, tell us at our enquiry form. Every report is read.
Sources
Retraction Watch database, published by Crossref (gitlab.com/crossref/retraction-watch-data) · Crossref and OpenAlex metadata · Problematic Paper Screener, Cabanac, Labbé and colleagues (dbrech.irit.fr) · Cabanac, Labbé and Magazinov, Tortured phrases: a dubious writing style emerging in science
, 2021 (arXiv:2107.06751).