Why You Should Re-learn How to Cite in the AI Era

When you think about citations in an essay, you might say: “Citations? Isn’t that something everyone already does?” Just read a lot of relevant literature, find content and insert it. But in an era when everyone is using AI-assisted writing and churning out drafts quickly, citations have become more important than ever.

My writing rhythm used to be: carefully read relevant literature → deliberate thinking about logic → start typing; but nowadays, almost everyone uses AI and writing speed has jumped dramatically, though model hallucinations have also produced a large amount of false content. Every sentence’s source, every argument’s evidence, may determine whether your paper will violate policy or be outright rejected.

I’ve seen too many friends casually using ChatGPT, auto-rewriting, polishing, yet overlooking two fatal questions:

1. Which sentence was my own? Which one did AI give me?
2. And for that AI-generated sentence, is there truly a piece of literature to back it up?

When AI writes so smoothly, we may forget: it won’t automatically give you reliable references. In fact, in many cases, the citations generated by AI are fake.

In other words: in the AI era, essay writing must adopt a new approach to citations. You can no longer use the old method of “meticulously read literature → slowly extract citations”; nor can you hand the whole writing task to AI and trust it to sort out citations for you. Instead, you should adopt a hybrid workflow: use AI to speed up your initial draft and structure, collect information; then you read and confirm the core arguments; and finally you use real academic literature to support each key claim.

Here is the specific process I commonly use:

• First, I ask ChatGPT to generate an outline to quickly map out the essay’s framework.
• Then I manually revise the outline and add logic.
• Next, I search for literature manually using keywords or using tools like Perplexity and GPT + web search to locate potentially useful articles.
• After confirming a piece of literature, I import it into NotebookLLM and use AI to quickly summarise the full text, extract key paragraphs, and identify page numbers or key tables.
• Finally, I download the article myself, read it in detail, pick out appropriate sentences or data to insert, then manually format the citation into my paper.

Based on the above process, here are two examples to help illustrate:

Example 1: The Impact of AI on Student Education

• Outline Generation
  • Prompt ChatGPT: “Please draft an outline for an essay on the impact of Artificial Intelligence (AI) on student learning outcomes in school environments, focusing on personalized learning, changes in teachers’ roles, students’ attention and self-learning abilities, and potential risks such as dependency or bias.”
  • After receiving the initial outline, manually revise and enrich the logic — for instance, by adding a section on “Policy and Ethics.”
• Locating Literature
  • Use keywords (e.g., “AI in education student learning outcomes personalized learning”) or search relevant articles through Perplexity or ChatGPT with browsing enabled.
• Importing and Summarizing with Tools
  • Import the selected papers into your notebookLLM or a similar tool to have AI quickly summarize the full text, extract key passages, and locate page numbers. For example, it might identify that “AI use reduces student stress” appears on page X, paragraph Y.
• Deep Reading and Citation Formatting
  • Personally download and carefully read the papers, selecting sentences or data suitable for your essay.
  • Then format the citation according to APA style.
• Reference entry (example): Smith, A. B., & Lee, C. D. (2023). Artificial intelligence in education: Effects on student autonomy and attention. Journal of Educational Technology, 12(1), 34–50. In-text citation: (Smith & Lee, 2023)
  • Insert into the essay where appropriate, specifying the page number if needed, e.g., (Smith & Lee, 2023, p. 38). Example 2: The Impact of AI on Employee Creativity
• Outline Generation
  • Prompt ChatGPT: “Please draft an outline for an essay on how Artificial Intelligence impacts employee creativity in the workplace, focusing on AI-assisted workflows, metacognitive ability, conditions for creative improvement, managerial and leadership roles, and possible negative effects.”
  • Revise the outline and supplement the logic.
• Locating Literature
  • Search using keywords such as “AI employee creativity field experiment generative AI” to find relevant articles.
• Importing and Summarizing with Tools
  • Import the selected papers into notebookLLM to extract key sections.
• Deep Reading and Citation Formatting
  • Download the full text, read deeply, and select useful sentences.
  • Format the citation in MLA style (example): Johnson, Mark L., and Emily R. Carter. “Generative Artificial Intelligence and Employee Creativity: A Field Experiment in a US Technology Firm.” Journal of Organizational Innovation, vol. 8, no. 2, 2024, pp. 112–130. In-text citation: (Johnson and Carter 118)

Compared with the traditional “read, write, cite” method, this workflow is much faster—but it has pain points: constant tool switching, back-and-forth searching, time spent verifying the relevance of literature, After repeated practice, I realised: the workflow works, but there are two bottlenecks—first, identifying truly relevant literature still takes time and effort, even when leads exist; second, the overall workflow is fragmented, so your speed gains get eaten by tool-switching and verification. So we started thinking: what if there were a tool that integrates “outline generation → literature lead search → precise paragraph locating → insertion of real citations”? Wouldn’t it save loads of time and risk?

We developed a tool that isn’t a “just plug AI in and you’re done” black box. Instead, it breaks the process down: AI generates the structure → automatically retrieves literature leads from real databases → you select them and quickly access the full text → the tool helps you mark up citations and export them. That way, every key claim is backed by a real, traceable reference, and the entire workflow can be done end-to-end in Knowee Writer, letting you write fast without sacrificing citation reliability.