Opening the book…
Citation and Documentation
The Mystery: Why Was the Paper Still Incomplete After the Student Disclosed AI Use?
Luis uses a generative AI tool while writing a research paper about college food insecurity. The tool helps him brainstorm keywords, reorganize two paragraphs, and summarize several findings.
At the end of the paper, Luis adds this sentence:
I used ChatGPT to help with this assignment.
He assumes that this statement solves every citation and academic-integrity concern. He does not cite the scholarly articles behind the AI-generated summary because he disclosed that he used AI. He also deletes his prompts and chat history.
His instructor returns the paper with questions:
Which claims came from scholarly research? What exactly did the AI contribute? Which version of the tool did you use? Did you verify the summary? Which words are yours?
Luis is confused. He did not hide his AI use, so what information is missing?
To solve the mystery, he must distinguish among citation, documentation, and disclosure. These practices overlap, but they do different jobs.
Citation, Documentation, and Disclosure Are Not Interchangeable
A citation connects a claim, quotation, idea, image, or other contribution to its source. It allows readers to determine where the information came from and, when possible, locate it themselves.
Documentation preserves information about the research and writing process. For AI-assisted work, documentation may include the tool, model or version, date, prompts, uploaded materials, generated responses, revisions, and verification steps.
A disclosure explains how AI contributed to the completed work.
A citation answers:
Where did this information or language come from?
Documentation answers:
What happened during the research and writing process?
Disclosure answers:
How did AI contribute to what the reader is seeing?
One statement may not answer all three questions.
Citation is a formal acknowledgment connecting information, language, media, or ideas to a source.
Documentation is a record of the tools, prompts, outputs, decisions, and verification steps used during a project.
Disclosure is a transparent explanation of how an AI tool contributed to the work.
An AI-use statement is a short disclosure naming the tool and describing its role in the project.
A prompt record preserves the instructions or questions supplied to an AI system.
Provenance is the history of where information came from and how it changed before reaching its current form.
Substantive use refers to AI assistance that meaningfully influences the ideas, language, analysis, evidence, design, or final product.
Citing an AI tool does not automatically disclose everything it did. Disclosing AI use does not replace citations to the real research sources.
Cite the Evidence, Not Merely the AI Summary
Suppose an AI tool tells Luis:
College food insecurity is associated with poorer health, greater academic difficulty, and lower graduation rates.
Even when the statement is accurate, the AI response is not the strongest evidence for it. Luis should locate and read the scholarly studies supporting those findings.
His paper should cite the research:
College food insecurity is repeatedly associated with poorer health and academic outcomes, although much of the evidence remains observational (Bruening et al.; Hagedorn-Hatfield et al.).
The AI may have helped Luis discover the idea or simplify the wording. The scholarly sources establish the evidence.
The Modern Language Association similarly distinguishes AI acknowledgment from ordinary source citation. Its current guidance explains that generative AI output may lack transparent provenance and recommends following generated claims to the original materials, reading them, and citing those sources directly.
A citation to ChatGPT does not turn ChatGPT into the author of the studies it summarized.
Disclosure Does Not Repair Missing Research
Luis’s sentence—“I used ChatGPT”—provides some transparency, but it does not identify which claims he verified or where the evidence came from.
Imagine that he writes:
Researchers have proved that campus food pantries raise graduation rates by 20%.
He adds an AI disclosure but provides no supporting study.
The disclosure tells readers that software may have helped produce the sentence. It does not establish that the statistic is real.
Luis must either locate credible evidence supporting the claim, revise it to match the available evidence, or remove it.
An AI disclosure is not a magical citation blanket that can be gently placed over unsupported information.
Citation May Be Necessary When AI Output Becomes Part of the Work
Citation or attribution may be appropriate when a student quotes, closely paraphrases, analyzes, reproduces, or substantially relies on AI-generated output.
For example, a paper studying AI errors might include a short generated response as primary-source evidence:
When asked for studies about campus food pantries, the system generated two references that could not be located in the named journals (ChatGPT, personal communication, July 18, 2026).
The exact formatting will depend on the required citation style, whether the output has a stable shareable link, and the instructor’s directions. APA, MLA, and Chicago continue to update their guidance as AI tools and their retrievability change. Current APA Style guidance provides specific reference formats for generative AI tools, while MLA advises writers to describe substantive AI use beyond the mechanics of a citation.
Students should therefore check the assignment, syllabus, institutional policy, and current style guide rather than copying an old citation example from an unrelated website.
A Citation Is Not Proof of Accuracy
Even a correctly formatted AI citation tells the reader only which tool produced the material.
It does not prove that the output was accurate, unbiased, or supported by evidence.
Consider:
ChatGPT reported that 62% of community college students experience food insecurity (ChatGPT, 2026).
The citation identifies the immediate source of the sentence. It does not identify where the percentage originated, which students were surveyed, or whether the number exists.
A stronger approach traces the statistic to the original study or report:
In the multicampus survey conducted by [verified authors], 62% of respondents at the participating institutions met the study’s definition of food insecurity ([verified citation]).
Now the reader can evaluate the population, methodology, and limits of the statistic.
AI citation documents the output. Source citation supports the factual claim.
What a Useful Disclosure Should Explain
“I used AI” is usually too vague to help the reader understand the work.
A useful disclosure identifies the tool, the purpose of the use, the parts of the process affected, and the human verification or revision that followed.
Compare these statements.
Too vague: I used AI for this paper.
More useful: I used ChatGPT in July 2026 to brainstorm database keywords and suggest possible ways to organize my verified research notes. I independently located and read all cited sources, checked the generated suggestions against those sources, and wrote the final analysis.
The second statement shows where the AI contribution begins and ends.
Scholars studying AI disclosure have argued that transparency supports appropriate credit, accountability, and evaluation of research practices. However, disclosure requirements differ across institutions, journals, disciplines, and types of AI use (Resnik & Hosseini).
One Disclosure Does Not Fit Every Project
The appropriate level of detail depends on how substantially the tool influenced the work.
A grammar checker that corrects a few typographical errors may require different documentation from a system that drafted paragraphs, analyzed data, generated code, translated interviews, created images, or reorganized an argument.
Kari Weaver’s Artificial Intelligence Disclosure Framework recommends describing AI use according to the role it performed within the research or writing process. The framework aims to make disclosures clear enough to show both the AI contribution and the human contribution (Weaver).
The central question is not merely:
Did I open an AI tool?
A better question is:
Did the tool meaningfully influence the ideas, language, evidence, analysis, organization, or final product that another person will evaluate?
Examples of AI-Use Disclosures
| Type of AI assistance | Example disclosure |
|---|---|
| Brainstorming and keywords | I used ChatGPT in July 2026 to generate possible keywords related to college food insecurity. I tested the terms in library databases and independently selected and evaluated all sources. |
| Organization | I used Claude in July 2026 to suggest possible groupings for research notes I had already collected. I revised the organization and wrote the final outline and paper. |
| Editing | I used Grammarly’s generative features to identify unclear sentences and suggest wording changes. I reviewed each suggestion and accepted, rejected, or revised it. |
| Drafted language | I used ChatGPT to produce an initial explanation of the difference between correlation and causation. I checked the explanation against assigned sources and substantially rewrote it before including it. |
| Image generation | I used an AI image generator to create the chapter illustration from an original prompt. No third-party image was supplied as a reference. The image is identified as AI generated in its caption and attribution. |
| No AI use | I did not use generative AI to develop, draft, revise, translate, or create media for this assignment. |
Figure 8.5.1 These are models, not universal rules. Students should adapt them to the assignment and disclose what actually occurred. (CC BY-NC; Cheryl Bailey)
Document the Process While You Work
Trying to reconstruct AI use after finishing a project is difficult.
Luis initially deletes his chats because he assumes they no longer matter. Later, he cannot remember which prompts produced which suggestions or whether the tool supplied wording that remains in the paper.
A better approach is to keep a short working record.
For each meaningful AI interaction, Luis records the tool and model if available, the date, the purpose, the prompt, the useful output, the changes he made, and the sources used for verification.
His note might read:
Tool and date: ChatGPT, July 18, 2026
Purpose: Generate alternative database keywords
Prompt: “Suggest scholarly search terms for research about food insecurity and degree completion among community college students.”
Useful terms retained: basic needs insecurity, degree attainment, academic persistence
Verification: Tested all terms in Academic Search Complete and ERIC
Final use: Search development only; no generated prose copied into the paper
This record makes the final disclosure easier and provides evidence of Luis’s research process.
Save Important Prompts and Outputs
A prompt affects the generated response. Recording only the tool name may therefore leave out a major part of the context.
Compare:
Tell me about student hunger.
with:
Identify possible scholarly search terms for studying the relationship between food insecurity and academic persistence among California community college students. Do not invent citations.
The prompts ask for different work and will likely produce different outputs.
When an instructor requires documentation, students may save a transcript, export the conversation, capture the relevant prompt and response, or place selected exchanges in an appendix. The MLA notes that writers may need to provide information beyond formal citation, such as notes describing AI use, prompts, or linked or appended transcripts.
Do not submit private, confidential, copyrighted, or personally identifiable information merely to prove that you used AI. Documentation should remain proportionate and safe.
Explain Human Responsibility
AI cannot accept academic responsibility for a submitted assignment.
The student decides which claims to include, which sources to trust, what language to submit, and whether the work meets the assignment requirements.
A disclosure should not sound like this:
ChatGPT wrote the inaccurate section, so the error was not mine.
A more responsible statement would explain the AI contribution while preserving human accountability:
I used ChatGPT to draft a preliminary explanation. I reviewed and revised the language and accept responsibility for the accuracy and final content.
Current scholarly guidance on AI disclosure emphasizes that authors must independently verify AI-assisted content and remain responsible for the completed work.
Disclosure explains the process. It does not transfer responsibility to the tool.
AI Is Not an Academic Author
Authorship involves more than generating words. An author accepts responsibility for the work, responds to questions, discloses conflicts, and stands behind the accuracy and integrity of the final product.
An AI tool cannot perform those responsibilities.
Therefore, students should not list an AI system as a coauthor of an academic paper. They may cite, attribute, document, or disclose its contribution according to the assignment and style guide.
The human author remains accountable.
Documentation Helps Distinguish Assistance From Substitution
An instructor may permit AI for brainstorming but not for drafting. Another may permit editing but require disclosure. A third may prohibit generative AI because the assignment specifically assesses independent writing or problem solving.
A process record can show how the tool was used.
For example, Luis’s saved prompts demonstrate that he requested keyword ideas and outline feedback. His document history shows that he drafted and revised the paper. His citations show that he located and read the scholarly sources.
Documentation does not automatically prove that every use complied with the assignment. It does make the process more transparent.
The safest approach is to read the policy before using the tool, not after submitting the assignment.
When AI Use Is Not Allowed
Citation and disclosure do not convert prohibited assistance into permitted assistance.
Suppose an instructor states:
Do not use generative AI to write or revise this reflection.
A student cannot submit AI-generated prose and avoid a violation by adding:
This reflection was generated by ChatGPT.
The disclosure makes the prohibited use visible. It does not make the use acceptable.
Citation addresses attribution. Compliance depends on the assignment rules.
When the policy is unclear, ask the instructor before using AI in a substantial way.
Copyright Attribution Is a Separate Question
An AI disclosure does not replace copyright attribution for third-party materials.
Suppose a student uploads a copyrighted illustration and asks an AI system to transform it into a different style. The final project may involve questions about the original image’s copyright, the permission to create a derivative work, and how the source should be attributed.
Likewise, asking AI to rewrite a copyrighted article does not remove the need to cite the article or create permission to reproduce protected expression.
Researchers must keep four questions separate:
What factual or scholarly source supports the claim?
What AI tool contributed to the work?
What disclosure does the assignment require?
What copyright or license applies to reused material?
One acknowledgment cannot perform all four jobs.
Example: Revising Luis’s Paper
Luis’s original paragraph says:
College food insecurity causes depression, lowers grades, and makes students drop out. ChatGPT says colleges should provide food pantries and emergency aid.
His only documentation is:
I used ChatGPT.
Luis revises the paragraph after locating scholarly sources:
Research consistently associates college food insecurity with poorer health and academic outcomes, although much of the evidence does not establish that food insecurity alone caused those outcomes (Bruening et al., 2017; Hagedorn-Hatfield et al., 2022). Campus food pantries and emergency assistance may address immediate needs, but additional longitudinal evaluation is needed to determine which interventions produce lasting improvements.
He then adds a disclosure:
AI-use disclosure: I used ChatGPT in July 2026 to brainstorm search terms and suggest a preliminary organization for this section. I independently located, read, and cited the scholarly sources. I checked all factual claims against those sources and wrote and revised the final prose.
The revised version separates the evidence from the AI assistance.
The scholarly citations support the findings. The disclosure explains the process. Luis’s notes document what happened.
Course Connection: Making the Research Process Visible
Citation helps readers follow ideas back to their sources. Disclosure helps readers understand where AI contributed. Documentation preserves enough of the process to explain and evaluate that contribution.
Together, these practices support transparency without pretending that AI output functions like an ordinary scholarly source.
The researcher must still locate credible evidence, verify claims, cite the original materials, and accept responsibility for the final work.
Solving the Mystery
Luis’s paper remained incomplete because the sentence “I used ChatGPT” did not explain what the tool did or identify the evidence supporting his claims.
His disclosure was too vague. He had not preserved his prompts or outputs. Most importantly, he had treated the AI-generated summary as a replacement for scholarly citations.
Luis solves the problem by separating three responsibilities:
Citation: He cites the scholarly studies supporting his claims.
Documentation: He keeps a record of the tool, date, prompts, outputs, revisions, and verification.
Disclosure: He explains how AI contributed to the submitted work.
His final disclosure states:
I used ChatGPT in July 2026 to brainstorm search terms and test possible organizational structures. I independently located and evaluated all cited sources, verified the factual claims, and wrote the final analysis.
The mystery was not whether Luis admitted using AI.
The mystery was whether his reader could determine what the AI contributed, where the evidence came from, and who accepted responsibility for the final work.
Check Your Understanding
How does an AI disclosure differ from a citation? Why should a student cite the original study instead of an AI summary of that study? What information makes a disclosure useful rather than vague? Why might a student preserve prompts and generated outputs? Does citing AI make a prohibited use acceptable? Who remains responsible for inaccuracies in AI-assisted work?
Citation identifies the source of information, language, or media. Documentation records the research and writing process. Disclosure explains how AI contributed.
AI-generated summaries do not replace citations to original scholarly or authoritative sources.
A correctly formatted AI citation identifies the tool or output but does not prove that its claims are accurate.
Useful disclosures name the tool, describe the purpose and extent of its use, and explain the human verification or revision that followed.
Students should preserve important prompts, outputs, dates, model information, and verification notes when an assignment requires process documentation.
AI should not be listed as an academic author because it cannot accept responsibility for the work.
Citation and disclosure do not make prohibited AI assistance acceptable.
The student remains responsible for the evidence, accuracy, originality, and final submitted work.
References
"AI References." APA Style, American Psychological Association, 2026.
"Beyond Citation: Describing AI Use in Your Work." MLA Style Center, Modern Language Association, 13 Aug. 2025.
Resnik, David B., and Mohammad Hosseini. "Disclosing Artificial Intelligence Use in Scientific Research and Publication: When Should Disclosure Be Mandatory, Optional, or Unnecessary?" Accountability in Research, 2025.
Weaver, Kari D. "The Artificial Intelligence Disclosure (AID) Framework: An Introduction." College & Research Libraries News, vol. 85, no. 10, 2024, pp. 407–11.
Attributions
No third-party AI conversations, student work, prompts, screenshots, or disclosure statements were reproduced. Luis’s mystery, research notes, sample paragraphs, disclosure models, and knowledge-check questions constitute original instructional content created for the LIB 15 OER.
The distinction between citation and broader acknowledgment draws on the Modern Language Association’s current AI guidance. The discussion of disclosure, transparency, and responsibility draws on Resnik and Hosseini and Weaver’s Artificial Intelligence Disclosure Framework.
This page titled 8.4: Citation and Documentation was last modified on Mon, 05 Oct 2026 17:12:15 GMT and is shared under a CC BY-NC-SA 4.0 license and was authored, remixed, and/or curated by Cheryl Bailey.
The book could not be loaded just now.