How to use #ChatGPT for Discussion Section in a #Research Study

A strong discussion section starts by turning research results into a clear narrative: summarize the main findings, restate the study’s objective and rationale, then interpret each hypothesis in relation to what’s already known and what theory predicts. The core workflow is to begin with a brief recap of the key results and why the study was worth doing, so the rest of the discussion has context. From there, each hypothesis gets its own treatment—describing what the study found, and then comparing and contrasting those findings with prior research to show where the work aligns, diverges, or extends earlier conclusions.

Artificial intelligence can help draft and organize this interpretation, but it must be used cautiously. Asking ChatGPT to compare results with previous studies is acceptable only as a starting point; the references it provides can’t be trusted. The fix is manual verification: search the relevant literature in Google Scholar (or other databases) and confirm that the cited papers actually exist and support the claims being made. Tools such as Elicit can also assist in locating sources, but the responsibility for accuracy remains with the researcher.

The next major ingredient is theory-based interpretation. Instead of stopping at “what happened,” the discussion should explain “why it happened” using an established framework from the literature. For example, when the relationship between corporate social responsibility (CSR) and organizational performance (OP) is the focus, ChatGPT can be prompted to connect the findings to the resource-based view (RBV). In that framing, CSR is treated as a source of valuable, hard-to-imitate resources—such as improved reputation, stronger stakeholder relationships, and other capabilities that are difficult to substitute. The goal is to translate empirical results into theoretically grounded mechanisms.

The same approach applies when results are not significant, which often requires extra care. If prior studies found a significant CSR–OP link but a new study finds insignificance, the discussion should still offer plausible explanations grounded in theory and existing research. Rather than blaming methods or analysis by default, the prompt can steer the model toward substantive reasons—again using RBV or the relevant theory—to explain why the expected relationship might not materialize in the specific context. Even when the reasons sound convincing, they still need literature support, which means returning to Google Scholar to check whether those theoretical explanations have support in published work.

Finally, the discussion should close with implications. ChatGPT can generate practical implications for significant and insignificant findings, but the output should be adapted to the study’s actual results and context. The emphasis throughout is on using AI as writing support—helping structure arguments, propose interpretations, and draft text—while ensuring every claim is accurate, referenced, and consistent with the researcher’s own reading of the literature.