Librarians face growing requests for AI-invented books

Librarians worldwide are increasingly receiving requests for books, magazines, and archival materials that do not exist, as readers ask for titles fabricated by artificial intelligence.

According to librarians, students and researchers regularly request sources that appear convincing at first glance but are not actually found in catalogs or archives. These are often fabricated scientific articles, non-existent journals, or fake archive numbers generated by chatbots such as ChatGPT, Gemini, or Copilot.

The International Committee of the Red Cross (ICRC), which oversees large library and archive collections, has also drawn attention to this problem. The organization emphasized that AI tools do not carry out research or check sources, but instead generate text using statistical language models. As a result, they can “invent catalog numbers, document descriptions, or even platforms that never existed.”

Sarah Falls, Researcher Engagement Chief  at Library of Virginia, told Scientific American that about 15% of the requests the library receives by email are texts generated by ChatGPT. According to her, this creates an additional burden for librarians, because proving that a unique document does not exist is sometimes more difficult than finding it.

Librarians and researchers also share similar experiences on social media. In particular, a scholarly communications specialist on the Bluesky platform said he repeatedly tried to verify links from a student’s reading list before discovering that they all originated from Google’s AI summary.

The ICRC highlights another fundamental flaw: AI is unable to honestly report that certain information does not exist, instead filling the gap with plausible but fictitious details.

In 2023, a tool was developed to recognize texts generated by artificial intelligence.

RELATED: Opening of the 2025 Frankfurt Book Fair: reflections on democracy, digital colonialism, and AI restrictions

Main image: Tag Hartman-Simkins / Futurism