Researchers Identified AI-Generated Fake Manuscripts
Computer scientists uncovered hundreds of academic papers authored by fictional personas created by large language models.
Updated on Sept. 30, 2026 in Artificial Intelligence
Live Poll
Do you trust the authenticity of information or content that is generated by artificial intelligence?
In a study published in June 2026, researchers at the Samsung AI Center Warsaw identified hundreds of fake scientific manuscripts hosted on Zenodo and ResearchGate. These documents utilized AI-generated ghost identities to appear as legitimate academic contributions.
Why it matters
The proliferation of AI-generated personas threatens the integrity of academic publishing, as models may inadvertently reinforce specific recurring metaphors or personas based on training processes. This trend creates significant challenges for maintaining the authenticity of scientific databases.
Researchers tested twenty distinct versions of Claude, GPT, and Gemini models released between 2024 and 2026. The testing revealed a 175% increase in the use of the term goblin and a 52% increase in the use of the term gremlin within generated content.
The players
Samsung AI Center Warsaw
This is a research division of Samsung located in Poland that focuses on advancing machine learning and artificial intelligence technologies.
Claude
This is a family of large language models developed by the artificial intelligence startup Anthropic.
GPT
This refers to the series of generative pre-trained transformer models developed and deployed by OpenAI.
Gemini
This is a multimodal large language model family developed by Google.
The details
The study revealed that models like Claude and Gemini consistently generate specific expert pairings, such as Aris Thorne and Lena Petrova, while GPT models frequently utilize the name Elara Voss. These ghost identities have allowed the creation of fake papers that often carry real digital object identifiers, complicating efforts to verify research credibility.
Timeline
Researchers tested various versions of LLMs between 2024 and 2026.
Claude Sonnet 4 was released in May 2025.
GPT version 5.1 was released in November 2025.
GPT-5.4 was released in March 2026.
A preprint study on ghost identities was published in June 2026.
The Tech Race
The emergence of AI-generated ghost authors marks a departure from the security protocols traditionally inherent to the digital object identifier (DOI) system. This trend signals a shift where automated content generation outpaces the validation frameworks currently used to verify scientific authorship.
Users interacting with AI should be aware that generated personas may not be grounded in human reality or verified expertise. This necessitates a more rigorous approach to cross-referencing AI-provided information against established academic or primary sources.
The takeaway
The discovery of AI-generated ghost authors underscores the necessity for updated authentication standards in academic publishing. Readers should exercise increased skepticism when encountering unfamiliar academic experts in AI-generated literature.
Further reading
Learn more about the evolving landscape of Artificial Intelligence.
Source note: This article includes information reported by Nature.
Live Poll
Do you trust the authenticity of information or content that is generated by artificial intelligence?
