The rapid integration of artificial intelligence (AI) technologies and large language models (LLMs) in the academic arena, without proper oversight or regulatory mechanisms, is creating new ethical challenges in scholarly work.
Experts warn that unmonitored AI use risks labor theft in research, undermines originality, and raises concerns about responsibility and reliability in scientific production.
The impact of AI on academia has become a growing topic of discussion, particularly regarding the potential ethical risks posed by uncontrolled usage in research and educational processes. While AI can assist with various academic tasks, unchecked reliance on these tools may erode critical human cognitive skills, including problem-solving, creativity and critical thinking, among students, educators and researchers alike.
Authorities and specialists are emphasizing the necessity of establishing clear ethical standards that allow AI to be leveraged effectively in academic work, such as theses, research papers, presentations and assignments, without replacing human intellectual labor or engaging in labor theft.
In a statement addressing these issues, professor Mehmet Serdar Güzel, director of the Ankara University Artificial Intelligence Institute, underlined that generative AI has become an indispensable tool in daily life, but its application in academia requires careful planning.
Güzel explained: “It is critical to define explicit ethical guidelines for where and how generative AI can be utilized in a thesis or academic research. Determining the appropriate scope and stages for AI use is essential to maintain academic integrity.”
He added that while AI can safely be employed for tasks such as formatting, language editing or summarizing text, it should not be relied upon for analysis, drawing conclusions or producing original content.
“Although texts generated by AI may appear unique, they are essentially compilations of existing material. Using AI without proper citation is considered plagiarism, and this must be acknowledged and addressed in academic planning. Producing work without proper references constitutes plagiarism, and using AI without attribution falls under the same risk,” Güzel noted.
Güzel further emphasized the continued necessity of human effort in critical academic activities such as research, literature review and drawing conclusions. “With the correct prompt engineering techniques, AI can help explore solutions, but these are not true research outcomes. We need to raise awareness among academics and students about this distinction. Unfortunately, misuse of AI in these contexts is already common,” he said.
The development of ethical standards for AI use is crucial both academically and societally. Güzel highlighted existing abuses, saying: “Some individuals use AI to generate fake data, while others use machine learning or AI algorithms to produce research results. There are even theses where both methodology and results are entirely AI-generated."
"We must put an end to this practice, but it is not straightforward. Programs like 'Turnitin AI' can detect some AI-generated text, yet AI continues to evolve, producing outputs increasingly similar to human work. We must change our mindset and recognize that generative AI can diminish students’ problem-solving abilities, creativity, and productivity,” he added.
Güzel suggested that AI could be integrated more effectively into academic workflows through human-machine interaction frameworks. “For example, producing a 400-page doctoral thesis may no longer be necessary."
"Often, 350 pages are reiterations of existing research. If AI can generate these sections efficiently, the actual contribution to science can be described in 10-15 pages, which may be more persuasive to reviewers. This approach could lead to a more efficient and valuable academic process,” he explained.
Current AI technologies are classified as “narrow AI,” meaning they simulate specific cognitive abilities rather than exhibiting human-level reasoning. Güzel elaborated: “What we see today is just the tip of the iceberg. In the future, general AI may be capable of reasoning, decision-making, and producing results similar to humans. By 2035 or 2050, machines might possess cognitive skills comparable to human reasoning. At that stage, clear rules and ethical frameworks defining the roles of humans versus machines will be essential.”
Professor Ebru Akçapınar Sezer from Hacettepe University’s Department of Computer Engineering emphasized that all participants in the publication process, including researchers, reviewers and journal editors, must ensure content is produced independently of AI models.
“LLMs can simplify certain tasks, but research dissemination and educational outputs must properly position AI tools,” Sezer said. “We are responsible for internalizing and applying rules that safeguard both ourselves and our professional standards. Misuse of LLMs can lead to academic outputs that lack authenticity and accountability.”
Sezer warned of scenarios in which fabricated research can be converted into publishable papers via AI. “One could submit a paper to a journal. Reviewers might use an LLM to simulate peer review, and the author could then use AI to respond. A paper could be accepted in this cycle, leaving humans with only command-writing roles. If both research production and validation are outsourced to AI, originality is lost, and multiple individuals effectively misrepresent work,” she explained.
While acknowledging that AI can assist in editing language and improving the presentation of scholarly work, Sezer cautioned that using AI to produce entire studies fosters a culture of academic complacency.
She also raised concerns about literacy and basic skill erosion, noting: “We may begin to see long-term consequences within five years. Just as reliance on calculators diminished certain arithmetic skills, AI could weaken reading and writing abilities, which are fundamental to knowledge creation. The short-term effects are already being debated, but generational skill erosion will become evident over time.”
