AI Hallucinations Flood Libraries as Chatbots Invent Academic Papers That Don't Exist

AI Hallucinations Create Crisis for Information Professionals

Librarians across institutions are facing an unprecedented challenge as AI hallucinations generate a flood of requests for materials that simply don't exist. Sarah Falls, chief of researcher engagement at the Library of Virginia, estimates that approximately 15 percent of all emailed reference questions her staff receives are now generated by AI chatbots like ChatGPT, many containing fake citations to imaginary sources1

1

2

2

. The burden on librarians has become so severe that Falls warns her team may need to limit the time spent verifying AI-generated fake content, fundamentally altering how libraries serve their communities1

1

.

The problem extends far beyond simple inconvenience. "For our staff, it is much harder to prove that a unique record doesn't exist," Falls explained3

3

. Students and researchers increasingly trust AI chatbots over trained information specialists, creating friction when librarians explain that requested materials are non-existent. This erosion of trust in human expertise represents a fundamental shift in how people approach academic research and information gathering2

2

.

International Organizations Sound Alarm on AI-Generated Fake Citations

The International Committee of the Red Cross issued a public warning about AI chatbots generating fake citations, cautioning that "generative AI tools always produce an answer, even when the historical sources are incomplete or silent"1

1

. The organization emphasized that these systems "do not conduct research, verify sources, or cross-check information" but instead generate content based on statistical patterns, producing invented catalogue numbers, document descriptions, and references to platforms that have never existed3

3

.

The ICRC's warning highlights a critical flaw in how AI chatbots operate: they cannot indicate when no information exists. Instead, they invent details that appear plausible but have no basis in archival records1

1

. This creates a specific threat to academic integrity as researchers may unknowingly build studies on fabricated foundations, perpetuating errors throughout scholarly journals and academic papers.

AI Inventing Academic Papers Contaminates Research Ecosystem

The crisis deepens as AI-generated fake content infiltrates professional scholarship. Andrew Heiss, an assistant professor at Georgia State University, discovered that AI hallucinations have created a cascading problem where fake citations appear in published articles, which are then cited by other papers, effectively laundering erroneous references4

4

. Each time he searched for bogus sources in Google Scholar, he found dozens of published articles relying on variations of the same made-up studies and scholarly journals4

4

.

This phenomenon of AI inventing academic papers creates particularly convincing fabrications because the citations often include names of living academics and titles resembling existing literature4

4

. The more these false citations are repeated from one article to the next, the more their authenticity appears reinforced, making disproving non-existent records increasingly difficult for librarians and researchers alike.

Real-World Impact on Academic Research and Information Access

The proliferation of AI slop is creating tangible problems across educational institutions. A scholarly communications librarian reported spending significant time looking up citations for a student, only to discover after finding zero results that the list came from Google's AI summary3

3

. High-profile cases include a Chicago Sun-Times freelance writer who generated a summer reading list with 15 books, ten of which didn't exist, and Robert F. Kennedy Jr.'s commission report that contained at least seven non-existent citations2

2

.

Researchers warn that authentic, human-written sources are being drowned out by the volume of fabricated content. One researcher lamented that "finding records that you KNOW exist but can't necessarily easily find without searching, has made finding real records that much harder"3

3

. This represents what cognitive scientist Iris van Rooij describes as nothing less than "the destruction of knowledge," as AI chatbots generating fake citations threaten to undermine our collective grasp on reliable data4

4

.

Why Users Trust Machines Over Human Experts

The phenomenon reveals troubling dynamics in how people interact with technology. AI chatbots speak in authoritative voices, leading users to trust them over trained librarians when conflicts arise2

2

. Some users even believe adding phrases like "don't hallucinate" to their prompts will ensure reliable output, though if such simple fixes worked, companies like OpenAI would already implement them automatically2

2

.

OpenAI released its agentic model for conducting "deep research" in February, claiming it operates at the level of a research analyst with lower hallucination rates than previous models. However, the company admitted the system struggles with separating "authoritative information from rumors" and conveying uncertainty when presenting information3

3

. Until more people understand the fallibility of generative AI, the burden will remain on human archivists to verify an ever-growing volume of reference questions involving fabricated materials1

1

. Librarians have served as critical thinkers and reliable information sources for thousands of years—unlike AI, they're trained to admit when they're wrong1

1

.