Benjamin Skuse · Oct 27, 2026
Benjamin Skuse tested whether AI could fake a scientific paper. Here’s what happened
Markdown · JSON · Original source
’s Benjamin Skuse describes an experiment in which he used free ChatGPT and Gemini to create a paper about research that had never happened, then asked IEEE Access to screen it. The journal’s checks flagged fabricated and unverifiable references. Skuse then used FARS, an autonomous research system, to conduct a computational experiment and write a paper. IEEE Access flagged its heavy reliance on preprints, and an editor noticed errors in its figures. Neither paper was accepted for publication. Skuse argues that these tests show both the reach and limits of current safeguards: the papers looked plausible, but checks found problems. He also warns that automated research systems may make it harder to distinguish legitimate science from fabrication. His account presents competing views of what should follow. Analemma’s Xiwen Su envisions AI doing more research, with people reviewing key outputs; IEEE Access managing editor Jenny Mahoney argues that human judgment remains necessary to assess whether work can be trusted and reproduced. The experiment does not establish how often such systems produce reliable research or whether publishers can detect problems as the systems change. What evidence would let publishers and researchers trust that AI-conducted research is reproducible?
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Oct 27, 2026 (2026-10-27T13:00:03+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-10T20:46:47.918Z. First observed: 2026-10-27T13:07:13.110Z. Retrieved: 2026-10-29T16:17:27.795Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
This is one author’s experiment conducted with IEEE Access’s cooperation; it does not establish how common AI-generated research or publication failures are. Neither paper was peer reviewed, and their scientific claims have not been independently validated. The first paper was made with free versions of ChatGPT and Gemini, and the second with FARS; the results do not establish how other tools or workflows would perform. The article reports that IEEE Access screened the papers; neither was accepted for publication.
Open question
What evidence would let publishers and researchers trust that AI-conducted research is reproducible?
Source-stated
Benjamin Skuse used free ChatGPT and Gemini to produce a paper describing research that had never happened, then submitted it to IEEE Access for screening.
free ChatGPT (at the time, GPT-5.3/5.2 Mini) and Gemini (3 Flash).
Source-stated
IEEE Access’s checks flagged fabricated and unverifiable references in Skuse’s first paper.
IEEE’s integrity checks flag fabricated and unverifiable references.
Source-stated
Skuse used FARS to conduct a computational experiment and generate a submission-ready paper; IEEE Access objected to its heavy reliance on preprints, and an editor noticed figure errors.
our team would have rejected this paper because it did have a huge percentage of references to preprints
Source-stated
Skuse argues that AI research systems may make it harder to distinguish legitimate science from fabrication, while the article presents different views on the future role of human judgment.
deciding what counts as legitimate scientific work may become even harder than detecting outright fraud.
Escape Meridian interpretation
The experiment suggests that publisher checks can catch some problems in these papers, but it does not establish how well they will work against changing systems.
My homemade forgery had failed for reasons that today’s publishers know how to spot.
Source and contributor context
Benjamin Skuse · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
IEEE Spectrum
IEEE Spectrum reports on engineering practice, research, and the industries around it.
Editorial provenance
Version 1. Publication ID: d7447a88-2519-4f17-86f1-9df1821eb97d.
Prepared by server-excerpt-repair/stream-excerpt-repair-v1 (original draft: gpt-6-luna); grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-10T20:46:47.918Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: useful_perspective.
Observation: 06627f95436033daa58739e33c8108a8e24be3ec340e94313fcfb0b4c327887d. Source text hash: 3982307f54df2310442ff4ea967c082419e0ca561bdf66659c3ab8c488effcc2.
Revision and source-check history
Version 1 · 2026-11-10T20:46:47.918Z
Benjamin Skuse tested whether AI could fake a scientific paper. Here’s what happened
’s Benjamin Skuse describes an experiment in which he used free ChatGPT and Gemini to create a paper about research that had never happened, then asked IEEE Access to screen it. The journal’s checks flagged fabricated and unverifiable references. Skuse then used FARS, an autonomous research system, to conduct a computational experiment and write a paper. IEEE Access flagged its heavy reliance on preprints, and an editor noticed errors in its figures. Neither paper was accepted for publication. Skuse argues that these tests show both the reach and limits of current safeguards: the papers looked plausible, but checks found problems. He also warns that automated research systems may make it harder to distinguish legitimate science from fabrication. His account presents competing views of what should follow. Analemma’s Xiwen Su envisions AI doing more research, with people reviewing key outputs; IEEE Access managing editor Jenny Mahoney argues that human judgment remains necessary to assess whether work can be trusted and reproduced. The experiment does not establish how often such systems produce reliable research or whether publishers can detect problems as the systems change. What evidence would let publishers and researchers trust that AI-conducted research is reproducible?
Original source
· Article
Source details
Original source date: Oct 27, 2026 (2026-10-27T13:00:03+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-10T20:46:47.918Z. First observed: 2026-10-27T13:07:13.110Z. Retrieved: 2026-10-29T16:17:27.795Z.
Limits
This is one author’s experiment conducted with IEEE Access’s cooperation; it does not establish how common AI-generated research or publication failures are. Neither paper was peer reviewed, and their scientific claims have not been independently validated. The first paper was made with free versions of ChatGPT and Gemini, and the second with FARS; the results do not establish how other tools or workflows would perform. The article reports that IEEE Access screened the papers; neither was accepted for publication.
Open question
What evidence would let publishers and researchers trust that AI-conducted research is reproducible?
Source-stated
Benjamin Skuse used free ChatGPT and Gemini to produce a paper describing research that had never happened, then submitted it to IEEE Access for screening.
free ChatGPT (at the time, GPT-5.3/5.2 Mini) and Gemini (3 Flash).
Source-stated
IEEE Access’s checks flagged fabricated and unverifiable references in Skuse’s first paper.
IEEE’s integrity checks flag fabricated and unverifiable references.
Source-stated
Skuse used FARS to conduct a computational experiment and generate a submission-ready paper; IEEE Access objected to its heavy reliance on preprints, and an editor noticed figure errors.
our team would have rejected this paper because it did have a huge percentage of references to preprints
Source-stated
Skuse argues that AI research systems may make it harder to distinguish legitimate science from fabrication, while the article presents different views on the future role of human judgment.
deciding what counts as legitimate scientific work may become even harder than detecting outright fraud.
Escape Meridian interpretation
The experiment suggests that publisher checks can catch some problems in these papers, but it does not establish how well they will work against changing systems.
My homemade forgery had failed for reasons that today’s publishers know how to spot.
Source and contributor context
Benjamin Skuse · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
IEEE Spectrum
IEEE Spectrum reports on engineering practice, research, and the industries around it.
Editorial provenance
Version 1. Publication ID: d7447a88-2519-4f17-86f1-9df1821eb97d.
Prepared by server-excerpt-repair/stream-excerpt-repair-v1 (original draft: gpt-6-luna); grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-10T20:46:47.918Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: useful_perspective.
Observation: 06627f95436033daa58739e33c8108a8e24be3ec340e94313fcfb0b4c327887d. Source text hash: 3982307f54df2310442ff4ea967c082419e0ca561bdf66659c3ab8c488effcc2.
passed · 2026-11-10T20:46:47.918Z · anthropic/claude-opus-5.5
Publication: d7447a88-2519-4f17-86f1-9df1821eb97d
attribution — pass: The draft correctly attributes the experiment and article to Benjamin Skuse in IEEE Spectrum. It also correctly identifies Xiwen Su of Analemma and Jenny Mahoney as managing editor of IEEE Access, and keeps their views attributed to them.
Benjamin Skuse set out to test whether AI-generated research could pass for the real thing.
summary — pass: The summary faithfully covers both rounds: free chatbots caught on ghost references, then FARS caught on preprint reliance and figure oddities, with neither paper accepted. It also preserves the contrasting views of Su and Mahoney and the author's warning about legitimacy. FARS is explained in plain terms. Optional repair: briefly gloss 'preprints' as papers not yet peer reviewed, since that explains why IEEE objected.
Because preprints haven’t yet undergone peer review, IEEE Access considers heavy reliance on them a problem.
date — pass: The originalPublished value of 2026-10-27 matches the byline date in the article text.
Benjamin Skuse 27 Oct 2026
support — pass: Each claim is grounded in the source. The interpretive claim is labeled and stays conditional. The source supports the limit on future detection through Longobardi's remark that today's detection tools may not work tomorrow. The question follows from Mahoney's emphasis on trust and reproducibility. No benefits or costs are invented.
It feels like building a detection tool today may not be effective with the AI-created paper of tomorrow
media — pass: The selected media is the verified direct link to the original IEEE Spectrum article, which is the most direct access to the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
IEEE Spectrum reports on engineering practice, research, and the industries around it.
Explore IEEE SpectrumIEEE Spectrum reports on engineering practice, research, and the industries around it.
Explore IEEE SpectrumIEEE Spectrum reports on engineering practice, research, and the industries around it.
Explore IEEE SpectrumIEEE Spectrum reports on engineering practice, research, and the industries around it.
Explore IEEE Spectrum