In recent months I have noticed a troubling pattern emerging across higher education: students relying on artificial intelligence tools are submitting work that cites authorities which simply do not exist. Fabricated journal articles, invented legal cases, and entirely fictitious quotations are appearing with increasing frequency. For those of us working in and around higher education law in England, this is no longer a hypothetical concern, it is an observable and growing risk with serious academic and legal implications.
At the centre of the issue is a phenomenon commonly referred to as AI hallucination. Unlike traditional research tools, AI systems do not retrieve sources in a verifiable way; they generate text based on patterns in data. While this can produce convincing prose, it can also result in authoritative-sounding but entirely false references. A case name may look plausible, a citation may appear correctly formatted, and a quotation may read persuasively, yet none of it exists in reality.
For students, the danger is obvious but often underestimated. In disciplines such as law, the accuracy of citations is not a peripheral concern; it is foundational. Legal argument depends on verifiable authority. Citing a non-existent case or misrepresenting a source is not merely a technical error, it strikes at the integrity of the work itself. Even where there is no intention to mislead, the submission of fabricated authorities is likely to be treated as academic misconduct.
This raises an important legal point: intent is fairly a factor when determining AI use. University regulations typically define misconduct to include the presentation of false or misleading information. A student who relies uncritically on AI-generated references may therefore find themselves subject to disciplinary proceedings, even if they did not realise the sources were fictitious. From an institutional perspective, the question is not whether the student meant to deceive, but whether the work submitted meets the required standards of accuracy and academic integrity.
There are also evidential implications. When concerns arise about the authenticity of sources, the burden often shifts to the student to demonstrate that their references are genuine and appropriately used. This can be particularly problematic where the student is unable to locate the original material because it never existed in the first place. What might have begun as an attempt to save time can quickly escalate into a formal academic misconduct case.
Beyond individual cases, there is a broader concern about the erosion of academic standards. If fabricated authorities go unnoticed, they risk contaminating the academic record and undermining trust in student work. For universities, this creates a reputational risk as well as a regulatory one. Institutions are expected to uphold rigorous standards of assessment, and failure to detect or address such issues may attract scrutiny.
There is also a professional dimension that should not be overlooked. Students in fields such as law, medicine, and finance are preparing for careers in which accuracy and reliability are essential. Developing a habit of relying on unverified information, however inadvertently, can have consequences that extend well beyond university. In legal practice in particular, citing a non-existent authority would be a grave professional error with potentially serious repercussions.
None of this is to suggest that AI tools have no place in higher education. They can be valuable aids for structuring ideas, summarising material, and exploring unfamiliar topics. However, they are not a substitute for proper research. Crucially, they should never be treated as authoritative sources in themselves.
The practical lesson for students is straightforward: every citation must be checked. If an AI tool suggests a case, article, or quotation, it should be independently verified. If it cannot be found, it should not be used. This is not simply good academic practice, it is essential risk management.
Ultimately, responsibility rests with the student. AI may generate the words, but it is the student who submits them. In an environment where the line between assistance and fabrication is increasingly blurred, diligence and verification are no longer optional; they are indispensable.
