Stop Before You Upload ! Is ChatGPT Saving Your Research Files and Causing Plagiarism?
Based on Dr Rizwana Mustafa's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
Use ChatGPT or Google AI Studio for editing suggestions by either uploading a file or copy-pasting text into the prompt.
Briefing
Using ChatGPT or Google AI Studio to edit a thesis or paper doesn’t automatically “save” a researcher’s file in a way that later triggers plagiarism or AI-detection problems, provided the work is handled through the tools’ normal upload or prompt workflows. The central takeaway is practical: researchers can request writing improvements—style, tone, and proofreading—without necessarily increasing AI or plagiarism scores in subsequent checks, as long as they use the tools to generate feedback rather than to outsource authorship.
The discussion starts with a common anxiety around data privacy and academic integrity: feeding a document into a large language model (LLM) raises fears that the text will be retained and reused, or that it will later be flagged by AI-detection systems as “not original.” To address that, the walkthrough focuses on how authors typically submit documents for revision. One method uses an upload feature (adding a file through the tool’s interface) and then asking for improvements to make the writing more academic and aligned with a target journal’s expectations. A second method involves copy-pasting the full text into the prompt area, then requesting section-by-section suggestions.
To test the concern, the creator runs two examples and compares AI/plagiarism results before and after using the LLMs. In the first case, a paper published in 2016 is used as a baseline. After uploading it to ChatGPT and Google AI Studio and requesting improvements to academic tone and writing style, the reported AI score remains at 0%, with the implication that the tool’s handling of the document did not raise the detection metrics. In the second case, a different paper—written with the help of AI—shows a much higher AI score (reported as 76%) while plagiarism remains minimal. That paper is then processed again through ChatGPT and Google AI Studio with prompts asking for improvements to each section, and the AI/plagiarism percentages are reported to stay the same.
From these comparisons, the conclusion is that using ChatGPT, Microsoft Copilot, or similar AI tools with either file upload or copy-paste does not inherently worsen AI-detection or plagiarism outcomes for the same document. The reasoning offered is that LLMs draw from patterns learned from online, publicly available sources (such as blogs, YouTube, and research papers), but they do not supposedly misuse the specific text a user uploads, nor do they provide that user’s content to other clients.
The practical guidance is to treat the AI output as editable suggestions. The results may be saved by the system, but the researcher can revise the text and even adjust prompts to reduce AI and plagiarism scores. The closing message points to follow-up content on using Google AI Studio or ChatGPT to improve academic tone while keeping AI-detection metrics from increasing, and on strategies to lower plagiarism using these tools.
Cornell Notes
The transcript addresses a key fear in academic writing: whether uploading a thesis or paper to ChatGPT or Google AI Studio causes later plagiarism or AI-detection issues. It describes two common workflows—uploading a file through the interface or copy-pasting the full text into the prompt—and then checking AI/plagiarism scores after processing. Two examples are used: a 2016 paper reportedly stays at 0% AI after revisions, while a separate AI-assisted paper reportedly remains at the same AI score (76%) with minimal plagiarism. The takeaway is that using LLMs for editing suggestions doesn’t automatically increase AI or plagiarism metrics for the same document, especially when the user keeps control of the final wording.
What are the two main ways the transcript says to use ChatGPT/LLMs with a research document?
How does the transcript test whether uploading affects AI or plagiarism scores?
What does the transcript claim about data privacy and whether uploaded text is reused?
Why does the transcript emphasize prompt changes and editing rather than accepting AI output as final?
What role do journal requirements and academic tone play in the workflow?
Review Questions
- What evidence does the transcript use to claim that uploading to ChatGPT or Google AI Studio doesn’t increase AI/plagiarism scores for the same document?
- How do the upload workflow and the copy-paste workflow differ, and what does the transcript say happens to AI/plagiarism results in each case?
- What strategies does the transcript suggest for reducing AI or plagiarism scores after using LLM feedback?
Key Points
- 1
Use ChatGPT or Google AI Studio for editing suggestions by either uploading a file or copy-pasting text into the prompt.
- 2
Request targeted improvements such as academic tone, writing style, proofreading, and section-by-section revisions.
- 3
Run an AI/plagiarism check before and after LLM-assisted editing to verify whether detection metrics change for your specific document.
- 4
Treat AI output as draft guidance and revise the wording yourself; don’t submit AI-generated text without review.
- 5
Adjust prompts and iterate on edits to potentially reduce AI and plagiarism scores.
- 6
Understand that LLMs learn from patterns in publicly available online material, but the transcript claims they don’t misuse user-uploaded documents or share them with other clients.