In July 2026, I sent signals about suspicious authors to the program committees of two conferences. In one paper, the researcher's name was not found in any scientific database; in the second, the email pointed to a well-known “paper mill.” A week later, the decisions came back: both papers were accepted. What's more—they were given oral presentations. This is what a failure of the scientific review system looks like.
What Is a “Fake Author”
There are several terms: “ghost author,” “guest author,” and “fabricated author.” All of these are cases where a person who did not participate in the research—or does not exist at all—is listed as an author. According to Retraction Watch, the number of retracted papers with authorship issues has grown by tens of percent over the past five years. This is especially noticeable in machine learning, where competition for publications is extreme.
Fake names are often used to “embellish” articles generated by neural networks. A fabricated name adds weight, helps pass peer review, and misleads readers. I call this scientific vibe coding: researchers generate papers via AI as easily as programmers generate code with autocomplete. The only difference is that code is easier to re-check than a scientist.
Case Study: Two Flags
First case: a paper on neural network architectures. The author was someone affiliated with an American university, but their email was on a free domain, and a search in OpenAlex returned only one odd publication, where the co-author was the same set of names. Second case: a paper on natural language processing. One of the co-authors had an ORCID created 24 hours before submission, and their publication history contained only this single preprint.
I collected evidence: screenshots, links to profiles, data from the Crossref API. I sent a flagging form to both conferences. The organizing committee's response was formal: “Thank you, we will check.” Judging by the subsequent decisions, the check came down to whether the author's name matched the declared topic.
How to Check Authors: 6 Steps
Here is the method that helped me find these anomalies. Any reviewer or reader can use it.
| Step | What to check | What to look for |
|---|---|---|
| 1 | ORCID | Whether profile exists, when it was created, how many publications |
| 2 | Databases | Scopus, Web of Science, OpenAlex: whether other papers exist |
| 3 | Affiliation | Whether the email domain matches the university's official website |
| 4 | Co-authors | Whether there are prior joint publications |
| 5 | Citations | Whether the authors are mentioned in other papers on the topic |
| 6 | Internet footprint | LinkedIn profile, Google Scholar, personal page |
For steps 2 and 5, it's convenient to use the Crossref API—it quickly returns metadata on all authors and their publications. I wrote a small Python script that compared the retrieved data. If there are zero matches, the probability of a fake, in my experience, is over 70%.
Why Both Papers Got Through
Classical peer review checks the scientific content, not the identities. Reviewers often see only the manuscript without links to profiles. Moderators are busy with format validation and plagiarism checks. An oral talk is a status given to roughly 5% of the best papers, but even it does not involve checking the author's biography.
The number of submissions to top conferences continues to grow. On major preprint archives, the number of new AI papers, by various estimates, has surpassed hundreds of thousands per year. Under such a load, platforms cannot cope with verification. Paper mills know this and exploit it.
How to Protect Yourself and Your Science
- For reviewers: before reviewing, check authors via ORCID and OpenAlex. It takes two minutes.
- For conference organizers: ensure that author verification is a mandatory step at submission. Use database APIs.
- For readers: if a paper looks suspicious, leave comments in open journals, flag the publication to Retraction Watch.
Automation is the key here. If you already build scientific data processing pipelines, build authorship checks into them. For example, through the Crossref API, as I did. ASI Biont supports Crossref connection via API—more at asibiont.com/courses.
Conclusion
My case is not a unique anomaly, but a symptom of a systemic shift. Vibe coding in science makes it possible to mass-produce plausible papers with fake authors, and existing defense mechanisms are not keeping up with this challenge. But if each of us—reviewer, editor, or simply reader—makes it a rule to check authors by at least two or three criteria, the situation will change. Fake authors will lose their main advantage: invisibility.
Comments