AI can help you map a field faster than you could on your own. It cannot replace the library. The citations it produces must be verified against the database before they appear in your paper — every time, without exception.
Where AI fits in the research workflow
Think of AI as the fastest orientation tool you have ever used. Ask it about the key debates in your topic, the major scholars, the dominant methodologies, the vocabulary of the field. Use that map to drive your search in your library's actual databases — which is where the real sources live.
What AI cannot do: give you a citation you can trust. Fabricated sources — plausible authors, realistic journals, credible titles that do not exist — are the single most common way students get into trouble with AI in research.
Source Tier · 20 XP
For a paper on the causes of the 2008 financial crisis, match each source to its tier.
Click through each to cycle: unknown → primary → secondary → AI-summary.
Primary sources are the original documents — the commission's own report, the participant's own account, the dataset itself. Secondary sources are analyses of those primary sources by scholars or journalists. AI summaries are not sources at all — they are uncited synthesis that may contain errors you cannot trace. Use AI to find your way in; then work in the tiers that count.
The library you are already paying for
Your university library subscribes to databases — JSTOR, PubMed, Westlaw, ProQuest, and dozens of subject-specific ones — that would cost thousands of dollars per year as individual subscriptions. Your tuition already covers them. A reference librarian can, in twenty minutes, teach you database search techniques that will save you hours on every paper you write for the rest of your degree.
The verification rule, reiterated
Every citation an AI generates must be independently confirmed in your library database or Google Scholar before it appears in academic work. Every statistic must trace to a primary source. Every factual claim that matters must be verified. This is not a suggestion. It is the basic obligation of anyone using AI tools in academic or professional work.
Citation Audit · 20 XP
An AI gave you this citation. What are your next three steps?
Fernandez, A. C., & Weiss, D. (2021). Credit default swaps and systemic risk in the post-Dodd-Frank era. Quarterly Journal of Financial Regulation, 38(2), 415–451.
Never ask an AI to verify its own output — it will often "confirm" a fabrication it just produced. Search the components independently in a database you trust. If the journal does not exist, the article does not exist, regardless of how real the authors sound. If the article is not where the AI says it is, you cannot cite it.