Get AI summaries of any video or article — Sign up free
Stop Before You Upload ! Is ChatGPT Saving Your Research Files and Causing Plagiarism? thumbnail

Stop Before You Upload ! Is ChatGPT Saving Your Research Files and Causing Plagiarism?

Dr Rizwana Mustafa·
4 min read

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.

TL;DR

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?

It describes (1) using the upload option—adding a file through the interface and then asking for improvements such as more academic writing style—and (2) copy-pasting the entire document into the prompt section and requesting suggestions, including section-by-section edits. In both workflows, the user then checks AI/plagiarism results afterward.

How does the transcript test whether uploading affects AI or plagiarism scores?

It compares reported AI and plagiarism percentages before and after running the same document through ChatGPT and Google AI Studio. One paper (published in 2016) is used as a baseline and is reported to show 0% AI after processing. A second paper written with AI is reported to show 76% AI and minimal plagiarism, and after processing again, the transcript claims the percentages remain the same.

What does the transcript claim about data privacy and whether uploaded text is reused?

It claims that LLM tools have data privacy protections and do not misuse the specific content a user uploads for their topic. It also asserts that the tools do not provide the user’s uploaded text or derived suggestions to other clients, even though the models are trained on patterns from online public sources.

Why does the transcript emphasize prompt changes and editing rather than accepting AI output as final?

It stresses that the AI’s output should be treated as editable suggestions. Because the user can modify the text and even “play with” prompts, AI and plagiarism scores can be reduced. The transcript also notes that results may be saved by the system, but the user can still change the final content.

What role do journal requirements and academic tone play in the workflow?

The transcript gives an example of asking the LLM to improve writing style to match the requirements of a Q1 category journal. The prompts focus on making the tone more academic and professional, rather than rewriting the work wholesale without review.

Review Questions

  1. 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?
  2. 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?
  3. What strategies does the transcript suggest for reducing AI or plagiarism scores after using LLM feedback?

Key Points

  1. 1

    Use ChatGPT or Google AI Studio for editing suggestions by either uploading a file or copy-pasting text into the prompt.

  2. 2

    Request targeted improvements such as academic tone, writing style, proofreading, and section-by-section revisions.

  3. 3

    Run an AI/plagiarism check before and after LLM-assisted editing to verify whether detection metrics change for your specific document.

  4. 4

    Treat AI output as draft guidance and revise the wording yourself; don’t submit AI-generated text without review.

  5. 5

    Adjust prompts and iterate on edits to potentially reduce AI and plagiarism scores.

  6. 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.

Highlights

A 2016 paper reportedly remains at 0% AI after being uploaded to ChatGPT and Google AI Studio for style improvements.
A separate paper reportedly shows 76% AI with minimal plagiarism, and the transcript claims those percentages stay the same after another round of LLM suggestions.
The transcript frames safe use as editable feedback: upload or paste the document, generate suggestions, then revise and re-check scores.
Prompt tweaking is presented as a lever for lowering AI and plagiarism metrics without abandoning the use of free AI tools.

Topics

Mentioned