Quality Control Checklist

Source Quality: Ensure at least 60% of sources are peer-reviewed (journals or conferences). Sources should span multiple perspectives rather than all saying the same thing. Sources should be recent unless historical context is needed. Be able to explain why sources were excluded if they were.

Citation Accuracy: Personally verify every citation by opening the paper and confirming the finding. Citations should be formatted consistently (choose APA, MLA, or Chicago and stick with it). No "mystery citations" are allowed—every claim must be sourced OR clearly marked as analysis. Include page numbers for all direct quotes.

Synthesis Quality: Identify clear themes across papers. Address contradictions rather than ignoring them. Note limitations of the research. Include at least 3 critical insights (not AI-generated). Ensure the synthesis actually answers the research question.

Writing Quality: The final document should read naturally (not obviously AI-generated). There should be logical flow from section to section. Use specific examples and data rather than vague generalities. Make it accessible to the intended audience.

Transparency: Be able to explain the search strategy if asked. Justify inclusion and exclusion criteria. Document where AI assistance was used. Verify AI outputs rather than blindly trusting them.

Academic Integrity: Do not copy-paste AI output without verification. Do not present AI-generated insights as original thinking. Properly attribute all ideas to sources. Be comfortable sharing the process with peers.

Intellectual Honesty: Note limitations and gaps in the literature. Acknowledge when evidence is mixed or uncertain. Distinguish between strong consensus and weak claims. Avoid cherry-picking studies that support a predetermined conclusion.

🚩 AI is probably hallucinating if: It cites a paper that was not uploaded or discussed. Citations include very specific page numbers that were not provided. Findings seem too perfectly aligned with the question. The quoted text cannot be found in the actual paper.

🚩 The synthesis is too AI-dependent if: The main findings cannot be explained without looking at the AI output. The methods of the studies being cited are not understood. Papers are being cited that have not been skimmed at all. What is novel or interesting about the findings cannot be articulated.

🚩 Moving too fast if: The research finished in under 30 minutes (unless it is a very narrow question). Not a single PDF has been opened. More than 50 sources are being cited (probably too broad or unfocused). The key papers cannot be remembered tomorrow.

Fix: Slow down. This is research, not a race.