Artificial intelligence can write, but it cannot be held accountable. A scientific text is a declaration of responsibility grounded in verifiable reality; when this declaration collapses, what remains is not science, but merely an imitation of it.

It is now nearly impossible to avoid using artificial intelligence in scientific work. Indeed, in many cases, it should be used. AI serves as a powerful aid to researchers in areas such as language editing, translation support, summarization, text planning, statistical coding, and the simplification of complex concepts.

First, we must seek the answer to this question: Does artificial intelligence support the researcher’s effort, or does it replace them by generating the text itself? There is a world of difference between a researcher using AI to better articulate their own ideas and having AI write the entire text, construct the arguments, and compile the bibliography, only to present it as one’s own academic output.

International publication ethics bodies have drawn this distinction. According to COPE, AI tools cannot be listed as authors because authorship entails more than just generating text; it involves taking responsibility for the accuracy and integrity of the work (https://publicationethics.org/guidance/cope-position/authorship-and-ai-tools). Similarly, the ICMJE emphasizes that the use of AI must be disclosed and that human authors bear full responsibility for any text, images, or sources generated by AI (https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-authors.html).

One of the most dangerous issues arising when this boundary is crossed is the phenomenon of AI “hallucinations.” Unsupervised models, in particular, are capable of generating books, articles, authors, journals, page numbers, and DOI details that appear authentic but do not actually exist.

A striking example of this from Turkey is a doctoral thesis titled “Political and Military Relations Between Golden Horde Khan Tokhtamysh and Chagatai Ruler Emir Timur,” prepared by Damira Makhanova at the Ankara University Institute of Social Sciences (https://tez.yok.gov.tr/UlusalTezMerkezi/TezGoster?key=KOgdn9H3uVnWeb15j2W4hxNPsKdqDnT_o7_nxp0jCES_jI7eqyjpDfhRGeNfUX8J). The thesis’s title and approval pages indicate that the work was conducted at Ankara University under the supervision of Prof. Dr. İlhan Erdem, that the jury included five professors, and that the defense date was set for January 27, 2026. On the ethics declaration page, the author states that all information was gathered in accordance with academic rules and ethical principles and that all sources were fully cited.

Certain publications attributed to İlker Evrim Binbaş are cited as sources in the thesis. In statements made on social media, Binbaş notes that some of these works are not his and describes the situation as serious (https://x.com/evrimbinbas/status/2059896658104549533?s=46&t=rFptNgTuGJR_LxxD3-c8gA). Of course, the presence of fake or unverifiable sources does not, in itself, conclusively prove that the text was written by artificial intelligence. However, this type of source usage strongly aligns with the problem of “hallucinated references” frequently encountered in the age of artificial intelligence.

More importantly, this case reveals a problem that runs deeper than just artificial intelligence. The academic oversight mechanism at Ankara University is failing to function. If a doctoral thesis systematically cites sources that do not exist or cannot be verified, this cannot be viewed merely as the student’s individual error. The supervision process, the jury evaluation, the institute’s checks, and the university’s quality assurance mechanisms must also be called into question.

This example holds symbolic significance in demonstrating just how fragile academic quality assurance has become in certain areas in Turkey. The issue is not simply the writing of a poor thesis, but the fact that such a thesis could pass through the academic review process. Even more alarming is that when such cases are dismissed as isolated errors, academic institutions end up perpetuating the mere production of documents rather than the pursuit of science.

It should be noted that this problem is not unique to Turkey. Cases of AI-generated fake citations are becoming increasingly visible in international academic publishing as well. Numerous fabricated sources have been identified even in articles published on academic ethics. One study indicates that one in every 277 articles indexed in PubMed contains fabricated sources (https://retractionwatch.com/category/by-reason-for-retraction/reference-problems/).

The fundamental issue is how to safeguard scientific integrity in the age of artificial intelligence. Artificial intelligence is a powerful tool when it supports scientific endeavor. However, when it transforms into a machine that generates text, arguments, and bibliographies without human oversight, it accelerates academic decay. A scientific text is a declaration of responsibility founded upon verifiable reality. When this declaration collapses, nothing remains but an imitation of science.