The best-documented fabrication scandals in journalism and science were caught the same way: someone checked a detail against a record the fabricator had not written. Janet Cooke invented a child. Stephen Glass invented a software company, then the notes and website to back it. Diederik Stapel invented the data for dozens of studies. Each invention satisfied editors, fact checkers or reviewers, and each fell to a college's files, a state's company register, a missing hotel receipt, or a dataset too tidy to be real.
EdCitation publishes this account of five cases from newsrooms and laboratories and one from a courtroom, each told from the published account of the outlet, court or inquiry involved, read on 24 September 2026, with the dates and figures those accounts give.
The lesson for a student is narrow: a source is real when its record says so, not when it looks right. That is the check EdCitation's free Verify references makes for every entry in a reference list.
Which fabrication scandals are best documented?
The six below, because in each the outlet, court or universities published their own account of what was invented and how it came out.
| Case | Exposed | What was invented | What exposed it |
|---|---|---|---|
| Janet Cooke, The Washington Post | April 1981 | An eight-year-old heroin addict, and parts of her CV | A prize biography that did not match the Toledo Blade's records |
| Stephen Glass, The New Republic | May 1998 | Quotations, sources, a company, and notes to back them | A Forbes reporter's search of tax and company records |
| Jan Hendrik Schön, Bell Labs | 2002 | Data in physics papers | Physicists who could not reproduce it; curves reused across figures |
| Jayson Blair, The New York Times | May 2003 | Comments, scenes, datelines; material lifted from others | A plagiarism complaint, then his own expense records |
| Diederik Stapel, three Dutch universities | 2011 | Datasets for psychology experiments | Three junior researchers' report to their head of department |
| Two lawyers, Mata v. Avianca | 2023 | Court decisions a chatbot produced | Opposing counsel and the court could not find them |
| A reference in your own paper | Before you hand it in | A source that does not exist, or was retracted | Verify references, free, no account |
Each invention was built to pass the check its author expected, and fell to a check against something the author did not control.
How were the journalism fabrications caught?
By testing a detail against an outside record: a college's files, a state's register of companies, an expense claim.
Janet Cooke and "Jimmy's World" (1980 to 1981)
On 28 September 1980 The Washington Post ran "Jimmy's World", a 2,256-word front-page story about an eight-year-old heroin addict. The city's police searched for the boy for 17 days and did not find him (Green, 1981). The story won the Pulitzer Prize for feature writing, announced on 13 April 1981.
The prize undid it. The Associated Press carried a biography, from a form Cooke had filled in for the nomination, giving her a Vassar degree and a Toledo master's. The Toledo Blade told the AP this did not match what it knew. Vassar's records showed one year of attendance, and there was no master's degree. Editors questioned Cooke on the night of 14 April, and the next day she wrote that the story was "in essence a fabrication" and resigned (Green, 1981). The prize was returned two days after it was awarded and went to Teresa Carpenter of The Village Voice (The Pulitzer Prizes, n.d.).
The Post's ombudsman, Bill Green, found that nobody at the paper had checked the form, and that her references had been checked only cursorily when she was hired. Green called it "a complete systems failure".
Stephen Glass and "Hack Heaven" (1998)
Glass joined The New Republic in September 1995 and began fabricating material in June 1996, according to the California Supreme Court, which reviewed the record in 2014 when it refused to admit him to the state bar (In re Glass, 2014). He testified that "all but a handful" of his 42 articles for the magazine contained fabrications; the editor, Charles Lane, had concluded in 1998 that 27 did. Glass also prepared notes and supporting material for the magazine's fact checkers.
"Hack Heaven", in May 1998, described a teenager hacking a California software company called Jukt Micronics. Adam Penenberg of Forbes Digital Tool tried to confirm it. Search engines found no website for the company; California's tax board said it had never paid taxes; the Secretary of State's office had no record of it; and Lexis-Nexis held one mention, Glass's own story (Penenberg, 1998).
Forbes telephoned Lane, who went with Glass to the Bethesda building where the story placed a hackers' convention; a person working there said none had been held. Glass then produced fake notes, business cards, a voicemail box, a website and newsletters, and had his brother pose as a source. Confronted with that, Glass admitted the story was invented, and Lane fired him (In re Glass, 2014).
Jayson Blair at The New York Times (2003)
On 11 May 2003 The New York Times published an investigation by five of its own journalists. Its reporter Jayson Blair had fabricated comments, concocted scenes, lifted material from other newspapers and wire services, and filed stories from places he had not been. They found new problems in at least 36 of the 73 articles he had written on national assignments since late October 2002 (Barry et al., 2003).
On 29 April Blair had been called in to answer a complaint of plagiarism from The San Antonio Express-News about a story datelined Los Fresnos, Texas. Asked to describe the house he said he had visited, he described it exactly; his editor learned later that the details could have come from the paper's own photo archive.
The records told the truth. Blair had filed from 20 cities in six states in five months without a single receipt for a hotel, rental car or plane ticket, and receipts he did file showed purchases in Brooklyn that he had placed in Washington. A naval medical centre whose patients he wrote about said its records showed he had never visited (Barry et al., 2003).
How were the science fabrications caught?
By researchers who looked hard at the data and found it impossible, or too good to be true.
Jan Hendrik Schön at Bell Labs (2002)
Schön's papers reported what the investigating committee called a remarkable number of major breakthroughs in condensed-matter physics. Other physicists could not reproduce his results, and by March 2002 the allegations about his data were public (Garisto, 2022). Bell Labs formed a committee in late May 2002, chaired by Malcolm Beasley of Stanford. By 20 June the allegations covered 25 papers and 20 coauthors; the committee examined 24 final allegations and, in its September 2002 report, found scientific misconduct in 16 (Beasley et al., 2002).
The evidence was in the figures. From the data files behind the plots, the committee found whole datasets, single curves and parts of curves reused to stand for different materials and devices, and precision beyond what a real experiment gives. In a paper on superconductivity in polythiophene, identical curves appeared several times in one figure, which the committee judged could not be real data. Virtually all the raw data files had been deleted. All of Schön's coauthors were cleared of misconduct.
A single retraction notice in Science on 1 November 2002 withdrew eight of his papers, according to Crossref's record of it, which draws on the Retraction Watch database (Bao et al., 2002). After the report he was fired, and he later lost his doctorate (Garisto, 2022).
Diederik Stapel at Tilburg, Groningen and Amsterdam (2011 to 2012)
The fraud came to light at the end of August 2011, when three young researchers in Tilburg University's social psychology department took their suspicions to the head of department after months of observation. Tilburg suspended Stapel in early September 2011 and dismissed him shortly after (Levelt Committee et al., 2012). Earlier warnings, from three other junior researchers and from two professors who had seen data "too good to be true", had not been acted on.
Three committees, one for each university where Stapel had worked, examined his entire body of work with teams of statisticians. Their joint report of 28 November 2012 found fraud established in 55 publications and evidence of fraud in 10 more, and fictitious data in chapters of 10 doctoral theses he had supervised. Sometimes questionnaires were never administered and Stapel supplied the data himself; sometimes real datasets were altered. The committees found no co-author or doctoral student complicit in the fraud.
The report also faulted reviewers and journals: hypotheses almost always confirmed, improbably large effects, impossible statistics, means identical to many decimal places, and almost none of it noticed. One of the 55 is a 2011 Science paper whose data, the report records, had supposedly been collected at a station; Science retracted it on 2 December 2011 (Stapel & Lindenberg, 2011).
What happened when lawyers cited cases a chatbot invented?
A federal judge in New York sanctioned them. On 1 March 2023 lawyers for Roberto Mata filed a submission citing court decisions that ChatGPT had produced. Avianca's lawyers replied on 15 March that they were unable to locate most of the case law cited, naming seven decisions; the court's own search failed too. Asked by one of the lawyers, ChatGPT had said its authorities were "real" and could be found in the legal databases. On 22 June 2023 the court imposed a $5,000 penalty and ordered letters to each judge falsely named as the author of a fake opinion (Mata v. Avianca, Inc., 2023).
Why AI tools invent references tells that story in full. The pattern is old: Glass built a company to survive a check, and a language model produces a citation shaped like a real one. The same check exposes both.
What do these cases teach about checking sources?
That a source is checked against a record its supplier did not write. Five rules follow, each from a case above:
- Check the record, not the notes. Glass's fact checkers checked notes he had written; Forbes checked the state's register. For a reference, the record is the publisher's page or an index such as Crossref, which Verify references searches for every entry.
- Never let a source vouch for itself. Cooke's biography came from a form she filled in; ChatGPT confirmed its own cases. A chatbot asked whether its reference is real is writing again, not looking up.
- Match every detail, not one. Blair's description of the house was perfect because it could be read off a photograph. A reference with a real author and journal can still be invented, so check title, authors, year and pages together, as How to check whether a reference is real sets out.
- Treat "too good to be true" as a question. Stapel's hypotheses were almost always confirmed; Schön's data were more precise than experiments give. A source that says exactly what your argument needs deserves a second look.
- Check for retraction, not only existence. Stapel's Science paper exists, and was retracted. Cite a source screens each source it builds for retraction, and our guide to retracted papers says what to do with one.
Which tool checks that a reference is real?
EdCitation's Verify references, and these cases show why it is the best tool for the job: every fabrication above survived the people who took the writer's word, and fell to someone who found the record. Verify references only looks. It searches for each entry in the publisher's record and never writes a reference, which no chatbot can promise; in Mata v. Avianca, the chatbot called its invented cases real.
Paste a list or upload the paper, free and with no account. Each entry comes back verified, "check this" or not found; a retracted paper is flagged; a source it could not check is shown as "could not check", never as not found. With the whole paper uploaded, each in-text citation is matched to the list and each entry back to the text.
Two retracted papers, run on 24 September 2026
We pasted two references into EdCitation's Verify references: the 2011 Stapel and Lindenberg paper with its DOI, and a 2000 Schön paper in Science on the fractional quantum Hall effect, without one. Both came back verified, because both papers exist, and both were flagged as retracted. The reasons, exactly as returned:
The DOI resolves to this record and the title matches. This paper has since been retracted.
Title, year and first author match a published record. This paper has since been retracted.
The summary counted them as two retracted papers, not two verified ones, which is the right emphasis: a real paper can still be one you must not rely on. Cite a source, given the Stapel DOI, built the entry from Crossref's record and listed the 2011 retraction notice beside it. Its APA 7 entry sets the title in sentence case; add the retraction in brackets, as our list below does.
To replace a source that fails, search the claim in Find sources, across about 300 million published works; replacing a weak or retracted source walks through it. References from a file, part of Pro at $8 a month (pricing), checks every reference in an uploaded paper and matches every in-text citation. None of these tools writes a word of a paper.
Quick questions
How was Stephen Glass caught?
Adam Penenberg of Forbes Digital Tool tried to confirm "Hack Heaven" in May 1998 and found no trace of its software company in tax records or California's company register. Glass later admitted the story was invented and was fired.
What did Jayson Blair fabricate?
Comments, scenes and datelines, and he lifted material from other publications. The New York Times's own investigation, published on 11 May 2003, found problems in at least 36 of his 73 national articles from late October 2002.
How many of Diederik Stapel's papers were fraudulent?
The three committees' joint report of November 2012 found fraud established in 55 publications and evidence of fraud in 10 more.
Did Janet Cooke give back the Pulitzer Prize?
Yes. The prize for "Jimmy's World" was returned two days after it was awarded in April 1981, once The Washington Post learned the story was fabricated, and went to Teresa Carpenter of The Village Voice.
How can I check that a source in my paper is real?
Look it up in the publisher's record rather than trusting how it looks. EdCitation's free Verify references does that for a whole list at once and flags retracted papers.
References
- Bao, Z., Batlogg, B., Berg, S., Dodabalapur, A., Haddon, R. C., Hwang, H., Kloc, C., Meng, H., & Schön, J. H. (2002). Retraction. Science, 298(5595), 961. https://doi.org/10.1126/science.298.5595.961b
- Barry, D., Barstow, D., Glater, J. D., Liptak, A., & Steinberg, J. (2003, May 11). Correcting the record; Times reporter who resigned leaves long trail of deception. The New York Times. https://www.nytimes.com/2003/05/11/us/correcting-the-record-times-reporter-who-resigned-leaves-long-trail-of-deception.html
- Beasley, M. R., Datta, S., Kogelnik, H., Kroemer, H., & Monroe, D. (2002). Report of the investigation committee on the possibility of scientific misconduct in the work of Hendrik Schön and coauthors. Bell Labs, Lucent Technologies. http://www.chem.ucla.edu/~craigim/pdfmanuals/misc/Lucent_researchreview.pdf
- Garisto, D. (2022, August 8). September 2002: Schön scandal report is released. APS News. https://www.aps.org/apsnews/2022/08/september-2002-schon-scandal-report
- Green, B. (1981, April 19). The players: It wasn't a game. The Washington Post. Archived August 27, 2017, at https://web.archive.org/web/20170827154429/https://www.washingtonpost.com/archive/politics/1981/04/19/the-players-it-wasnt-a-game/545f7157-5228-47b6-8959-fcfcfa8f08eb/
- In re Glass, 58 Cal. 4th 500 (2014). https://www.calbar.ca.gov/sites/default/files/agenda/agendaitem1000024727.pdf
- Levelt Committee, Noort Committee, & Drenth Committee. (2012). Flawed science: The fraudulent research practices of social psychologist Diederik Stapel. https://www.rug.nl/about-ug/latest-news/news/archief2012/nieuwsberichten/stapel-eindrapport-eng.pdf
- Mata v. Avianca, Inc., No. 22-cv-1461 (PKC) (S.D.N.Y. June 22, 2023). Opinion and order on sanctions. https://storage.courtlistener.com/recap/gov.uscourts.nysd.575368/gov.uscourts.nysd.575368.54.0_3.pdf
- Penenberg, A. L. (1998, May 11). Lies, damn lies and fiction. Forbes. https://www.forbes.com/1998/05/11/otw3.html
- The Pulitzer Prizes. (n.d.). Teresa Carpenter. https://www.pulitzer.org/winners/teresa-carpenter
- Stapel, D. A., & Lindenberg, S. (2011). Coping with chaos: How disordered contexts promote stereotyping and discrimination. Science, 332(6026), 251–253. https://doi.org/10.1126/science.1201068 (Retraction published 2011, Science, 334[6060], 1202)