Why ChatGPT Should Not Be Used for Academic Research

ChatGPT can speed up parts of academic work—summarizing articles and drafting text—but its outputs are unreliable when they’re used as sources for research claims, especially citations. A key warning emerges from an example on “abductive reasoning” linking Green HRM to organizational identification: ChatGPT can produce a plausible-sounding chain of logic, yet that doesn’t guarantee the underlying evidence exists or is correctly attributed.

In the example, Green HRM is treated as the integration of environmental sustainability principles into human resource management practices, while organizational identification is defined as the degree to which employees identify with and feel committed to their organization. From there, ChatGPT generates a reasoning pathway: Green HRM practices may increase organizational identification because sustainability initiatives can align with employees’ personal values and beliefs, strengthening their sense of connection and commitment. The response also suggests a secondary mechanism—green HR practices can boost employee engagement and motivation, which can further support identification.

The problem appears when the conversation shifts from generating reasoning to supplying academic references. When asked for citations to support the claim, ChatGPT provides specific author-and-year references (including “Khan et al 2016” and “Gang et al 2018”). But a follow-up check using Google Scholar finds that at least one of those citations cannot be located in the scholarly index. That mismatch is treated as a concrete sign that citation lists produced by ChatGPT may be fabricated, inaccurate, or otherwise not verifiable.

The takeaway is practical: even if ChatGPT produces a well-structured argument that can be edited and reused, citations generated by the system should not be accepted at face value. For literature reviews, the safer workflow is to search and verify using established academic databases and the original research articles themselves, rather than relying on ChatGPT’s reference suggestions. The guidance is aimed at keeping researchers “alert” during use—benefiting from drafting and summarization while maintaining strict verification standards for evidence.

Overall, the core finding is a reliability gap: ChatGPT can help with writing and reasoning drafts, but it cannot be trusted as a source generator for academic literature. That distinction matters because literature reviews and evidence-based claims depend on traceable, checkable scholarship; a single incorrect citation can undermine credibility and derail the research process.