Invisible Sources: Transparency and Attribution as Foundations of AI Literacy Instruction

Integrating AI attribution and transparency in library instruction

Posted on November 10, 2025 in Blog Posts

Generative AI creates both convenience and critical reflection within higher education. Tools like ChatGPT or Claude promise quick answers and creative shortcuts, but they often obscure the very processes that make them work. Their outputs are polished but opaque: little to no citations, a hard-to-follow evidence trail, and no clear sense of the voices and data that shape them. For students, this creates an illusion of authority. Without explicit guidance from trusted sources like librarians, students can accept AI-generated content at face value, unknowingly reproduce bias, or intentionally or unintentionally present work shaped by AI without acknowledgment.

As scholars, however, we know that knowledge does not appear from nowhere. Isaac Newton’s phrase “standing on the shoulders of giants,” later adopted as Google Scholar’s tagline, captures the cumulative and collaborative nature of academic work. Our ideas build on what came before, and acknowledging those foundations is an act of both respect and intellectual honesty. When students use AI without attribution or transparency for a variety of different reasons, that chain of lineage is broken.

This is where librarians play a vital role. By teaching students to note the invisible sources in AI outputs—that is, by recognizing both what is present and what is missing—we can help them approach generative tools with a critical lens, grounded in transparency, attribution, and respect for the collective nature of knowledge.

The Transparency Gap in AI

Transparency and attribution of information sources have always ensured the chain of knowledge remains visible, accountable, and connected across decades. Generative AI unsettles this tradition. When students use AI tools without disclosure, whether because they weren’t allowed to use them for a specific class or they don’t know how to cite them, they obscure the role of these systems in shaping their work. When they cite unverifiable AI outputs or fail to question the origins of an output, they remove their scholarship from the lineage of sources and perspectives that give academic writing its credibility. Hidden AI use is not simply a technical issue or an academic integrity issue; it is a break in the long-held scholarly contract.

This problem has direct pedagogical implications. Librarians have long taught students to interrogate the sources of information: Who created this? In what context? For what purpose? Generative AI resists such interrogation by design, concealing the very features that enable evaluation. One exercise I’ve tried with classes to expose this gap in practice is through a verification exercise: asking students to verify each claim in an AI-generated paragraph about a topic. Framing AI outputs as starting points that need to be proven rather than end points equips students to treat AI-generated text with appropriate skepticism.

Our instructional work already emphasizes the evaluation and ethical use of information. Extending these principles to AI is a natural progression of information literacy skills. By foregrounding transparency (disclosing how AI has been used) and attribution (acknowledging whose work underpins ideas, whether human or machine), we reinforce academic integrity and help students understand their place within the broader world of knowledge creation.

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• Enhancing Reference Interactions with Scopus AI
• How to Interview Your Library’s Patrons
• Beyond the Job Title: Data Strategist

📹 Register now for our upcoming webinar, “‘Data’ and Indigenous Studies: Entry Points for Academic Libraries”

---

The Librarian’s Role: Teaching Attribution & Transparency within Instruction

Librarians are uniquely positioned to address the instructional challenges posed by generative AI, whether as part of generalized information literacy instruction or AI Literacy instruction. Our expertise in citation, evaluation, and source literacy directly aligns with the questions students must ask of AI outputs, just as with any other information source: Where does this information come from? How can it be verified? What assumptions is it making? On my own campus, conversations with faculty and administrators are positioning librarians as trusted voices on AI literacy, giving us both opportunity and responsibility in shaping student practice.

Along with the exercise in the previous section, I’ve also integrated these types of exercises into one-shot instructions:

• Bias Analysis: Asking students to examine omissions, emotional tone, or framing in AI-generated text. This draws parallels to how we examine our own positionality as researchers.
  
  
• Disclosure & Attribution Examples: Providing sample statements and asking students how to make them better or more specific, and then having students draft their own. This normalizes AI transparency as part of academic integrity. Some institutions, whether at a course or faculty level, already require disclosure statements, and I use those as examples. Sharing these emerging practices in instruction demonstrates to students that transparency is not an optional courtesy but a professional norm in development.
  
  
• Comparative Reading: Ask students to analyze the differences between an AI-generated response and a scholarly source on the same topic. This helps surface what is gained or lost in each. I have also asked students to select a peer-reviewed article and give the AI tool the same title and abstract to write the article on, rather than a general topic.

These activities help students understand that transparency is a shared responsibility between scholars and AI knowledge creation. It is important to distinguish this instructional work from important structural frameworks such as the AID model, which proposes systemic approaches to disclosure in publishing or research. While such models are valuable, teaching transparency and attribution with these types of activities demonstrates the need for habits of mind that create critical reflexes to carry into varied academic contexts. They support equipping students to critically engage with AI outputs responsibly within the scholarly ecosystem.

Conclusion

At its core, AI literacy is not only about learning to navigate new technologies, but also about preserving the traditions that sustain academic life. Transparency and attribution ensure that students remain connected to the scholarly continuum, where knowledge builds off one another and contributions are recognized. In a moment when generative AI hides origins and erases lineages, these practices have become even more essential to the scholarly record.

Librarians are central to this work. By embedding lessons in transparency and attribution into AI literacy instruction, and larger information literacy sessions, we equip students to carry forward values that have long defined academic inquiry. We can model how curiosity, skepticism, and acknowledgment can coexist with technological innovation.

*

This post was written with the use of ChatGPT, with the prompt “act as an editor for clarity and conciseness” on original author text. All content was reviewed, edited, and refined by Bronte Chiang, who bears full responsibility for its accuracy and originality.

Related Posts