Skip to content
A network of connections on a gradient background.
Clemson University

Writing Lab

Make An Appointment

AI in Academic Writing

Overview

Here's the thing: generative AI writing tools are just that...tools. Tools, when used correctly and with good intentions, are helpful. Tools, when misused and abused, are harmful and ineffective.

AI tools can can save time, help with the clarity of your work, and give you valuable insights into effective communication. However, AI tools can also provide false information, misrepresent material, and violate academic integrity.

The term "generative AI" refers to AI tools that create things like writing, pictures, videos, and music. These generative AI models create these things based on databases from which they "learn" the information. 

We are at a turning point in writing, in academic, professional, and creative contexts. This particular page will focus on use of generative AI in academic contexts. 

Are all generative AI tools bad? Is every use of generative AI bad? No. Use of a tool is always context-bound. For example, a hammer, when used wrong or in the wrong situation, can lead to injury or something breaking. Like all tools, AI must be carefully harnessed and engaged. 

What is generative AI?

Generative artificial intelligence (AI) refers to models or algorithms that use databases to identify patterns and structures to generate content. That content can be text, images, audio, etc. This allows users to provide the AI with prompts or requests. For textual generation, AI uses Large Language Models (LLMs) to analyze vast amounts of textual data. This data is then used to create a neural network, allowing the model to mimic human response and various writing styles. 

The textual data used to "train" the LLMs is often drawn from the internet, but it is not cited or tracked where the data comes from or who created the original text. In general, these generative AI LLMs cannot determine whether or not text is biased or unbiased, accurate or inaccurate, or wholly incorrect. While the training of AI models may improve with time, the possibility of biased or inaccurate generation makes it incredibly important that users be mindful and critical of AI usage. 

There are many generative AI tools. Some are free and some are at a cost. New generative AI tools are being developed as time passes. If you sign up for any version, be aware of their terms and conditions, particularly regarding personal data and privacy. 

Here are some well-known generative AIs for writing:

  • ChatGPT
  • Google Bard
  • Microsoft Copilot
  • scite.AI
  • Wordtune
  • Quillbot

Benefits & Potential Uses

Generative AI, with its capacity to simulate human-like creativity and intelligence, offers a myriad of benefits across diverse domains. Firstly, it revolutionizes content creation by automatically generating high-quality written and visual material, facilitating efficient and scalable production. In the realm of art and design, generative AI becomes a powerful tool for creating unique and visually captivating graphics. Additionally, it significantly contributes to the development of chatbots and virtual assistants, enhancing user interactions through natural language generation. Language translation services also benefit from generative AI, providing more contextually relevant and fluent translations. Lastly, in personalized content recommendations, generative models analyze user behavior to offer tailored suggestions, optimizing user engagement. While these applications showcase the positive impact of generative AI, ethical considerations and responsible implementation are crucial to mitigate potential pitfalls and risks associated with its use.

  1. Content Creation: Generative AI can be used to automatically generate high-quality content, including articles, blog posts, and creative writing. This is particularly useful for individuals and businesses looking to produce large volumes of content efficiently.

  2. Creative Design and Art: In the realm of visual arts, generative AI, such as style transfer algorithms, can be employed to create unique and aesthetically pleasing images. Artists and designers use these tools to explore novel artistic styles and generate visually stunning graphics.

  3. Chatbots and Virtual Assistants: Generative AI plays a crucial role in the development of chatbots and virtual assistants. Natural language generation models enable these systems to respond to user queries and provide information in a conversational and human-like manner, improving user engagement and satisfaction.

  4. Language Translation: Generative AI models can enhance language translation services by generating more contextually relevant and fluent translations. This is particularly beneficial for translating complex or idiomatic expressions, improving the overall quality and accuracy of language translation applications.

  5. Personalized Content Recommendations: Generative models can analyze user preferences and behaviors to generate personalized content recommendations. This is widely used in recommendation systems for platforms like streaming services, e-commerce websites, and social media, providing users with content tailored to their individual interests.

Drawbacks & Risks

Generative AI, while promising transformative capabilities, introduces several pitfalls and risks that necessitate careful consideration. Foremost among these concerns is the potential for bias, as models may inadvertently perpetuate or exacerbate biases present in training data, raising ethical challenges in sensitive domains. The creation of deepfakes and realistic fake content poses a serious risk to the spread of misinformation, with implications for societal and political consequences. Security risks emerge as malicious actors exploit generative AI to craft convincing phishing emails or forged documents, undermining cybersecurity. The lack of accountability in tracing the origin of generated content further complicates efforts to address malicious uses. Lastly, overreliance on automation raises the risk of diminished human oversight, creativity, and judgment, emphasizing the need for a balanced integration of generative AI to harness its benefits while mitigating potential drawbacks. A thoughtful approach, incorporating technological safeguards and ethical guidelines, is essential to navigate these challenges effectively.

  1. Bias and Ethical Concerns: Generative AI models can inadvertently perpetuate and amplify biases present in the training data. If the training data contains biases, the generated content may reflect or even exacerbate those biases. This raises ethical concerns, particularly in sensitive domains like healthcare, finance, and criminal justice.

  2. Misinformation and Deepfakes: Generative AI can be used to create realistic fake content, such as deepfake videos or text that mimics a particular writing style. This poses a significant risk for spreading misinformation, as the generated content may be indistinguishable from authentic content, leading to potential social and political consequences.

  3. Security Risks: Malicious actors could use generative AI to create convincing phishing emails, fake news articles, or forged documents. This poses a threat to cybersecurity, as people may be more likely to trust and act upon content generated by AI systems.

  4. Lack of Accountability: With the rapid advancement of generative AI, it becomes challenging to trace the origin of generated content. This lack of accountability can be exploited for malicious purposes, as it may become difficult to attribute the creation of harmful or fraudulent content to a specific source.

  5. Overreliance on Automation: Depending too heavily on generative AI for content creation or decision-making can lead to overreliance on technology. This may result in a diminished emphasis on critical human oversight, creativity, and judgment, potentially impacting the quality and authenticity of work.

The two previous sections were written entirely by ChatGPT 3.5 with the following prompts: 

Could you provide five benefits of generative AI usage?

Could you provide five pitfalls and risks of generative AI usage? 

Could you write two separate 150-word introductions for the previous two prompts?

While these prompts generated fairly helpful responses, the AI does not understand or know the context in which this writing will be used. Therefore, the examples are not particularly focused on writing, communitication, or academia. Because the prompt included no information regarding the intended audience, the AI could not search for and generate for than audience. 

Likewise, notice that the AI does not indicate where it found this information. All of the above points were found through its database, but it cannot specify who wrote the text it trawled to learn. 

Also, note the tone of the two sections. Do they match the tone of the introduction? Most often, generative AI is recognizable for its formality or unnatural tonality. Even the slightest shift is recognizable. 

Policies Regarding AI Use

When using AI for writing support, it is important to note that many institutions outright prohibit AI-created content. Authorship guidelines are changing every day, so you must be aware of your particular context. For example: 

  • Springer Nature prohibits use of AI to generate images for manuscripts. Texts generated by AI should be clearly stated, and the AI will not be considered an author. 
  • Science journals prohibit the use of AI to generate text. Images and figures generated by AI can be used only with editor permission.
  • JAMA and the JAMA network journals allow use of AI generated content or assistance in writing but use must be reported in the manuscript. The AI will not be considered an author.
  • The International Conference on Machine Learning prohibits content generated by Generative AI, unless it is part of the research study being described.
  • Elsevier allows the use of AI and AI-assisted writing, but the AI should not be listed as an author. Images and figures generated by AI are prohibited.

So, with all of these varying perspectives on allowance and prohibition, it can be very difficult to gauge when you CAN use AI and when you CAN'T. When in doubt, verify. 

Email your instructor, the editors of the journal in which you are seeking publication, and speak to peers and colleagues before using generative AI content.

While Clemson does not have a policy specifically regarding the usage of generative AI, most instructors default to the academic integrity statement. Most essentially, the newest revision of the Clemson University academic integrity statement adds, "using materials generated using artificial intelligence that are turned in without attribution is considered plagiarism." 

The Writing Lab's policy aligns with the University's. When working with you to provide feedback, consultants will voice their concerns if writing is AI-generated and encourage you to make sure that is attributed. We would do the same if we are concerned regarding any form of plagiarism. If your instructor encourages use of generative AI, or has different policies on its usage, please inform your consultant. 

The most important takeaway? Be aware of the risks when using generative AI. Different contexts have different policies. As circumstances change and conversations move forward, these policies may be revised or changed. It is your responsibility as the writer to make the best decisions on behalf of yourself and those you represent (such as your lab, company, or team).

AI as Writing Assistant

Generative AI can be useful in a number of ways. It can help to think of writing as a process rather than a product. Often, writers feel very isolated in a writing process. Likewise, many writers struggle with positive perceptions of writing. That is to say, it's their least favorite aspect of any work that they do. So, writing the [insert genre here] can be the most anxiety-inducing or the most avoided. In these cases, having some kind of assistance can help the process move forward. Unfortunately, using AI to write the ENTIRE [insert genre here] will not quite work for most circumstances. Either it will get you into trouble, per policies and rules and general perceptions, or it will ineffectively address the needs of your context. 

Generative AI is a tool that can be used in various ways beyond writing your whole [insert genre here]. 

Let's follow it through a writing process.

Brainstorming

  • Generative AI can aid in finding ideas for your projects.
    • For example, if you give ChatGPT a prompt for "give me ideas for a place-based rhetorical analysis at Clemson University," the AI provides ideas for analysis of academic spaces and athletic facilities among other potential avenues. While this is likely too vague for an actual assignment, it could help in choosing a topic. 
  • Generative AI can help in narrowing down your topic. 
    • For example, if you give ChatGPT a prompt like "I want to do a rhetorical analysis of the library at Clemson. Could you help me narrow that topic down?", it will generate a list of "potential angles." These possible angles, such as architectural design and branding, may help you to be more specific in your analysis. 

Researching

  • Generative AI can help when developing a list of search terms for research.
    • For example, ChatGPT can help you in brainstorming search terms for research. If you give it the prompt, "Could you develop a list of five search terms for a paper on Indigenous history at Clemson?", it will list search terms like "Land Acknowledgement Clemson University" and "Clemson University Indigenous Initiatives." 
  • Generative AI can help you to better understand and track research.
    • For example, scite.AI will compile research within the databases it has access to using its best understanding of your prompt. While it is not always wholly accurate, it can help in seeing how topics can interconnect. 
    • Research Rabbit, Aria, a plug-in for Zotero, can create concept maps to so how sources connect and how patterns emerge across papers. Elicit does the same thing, which can be very helpful.
    • When ChatGPT Pro is used as a plug-in, it can create word clouds and network-like maps. 
  • Generative AI can help in formatting references (with some revisions required). 
    • Platforms like EndNote, Zotero, and RefWorks can use algorithms to read metadata and format citations in various styles. Proceed with caution, some or most of the time, the metadata is incorrect. Citation generators can help, but you must ALWAYS check to make sure they are accurate.
  • Generative AI can help with research summaries.
    • Platforms like scite.AI, Elicit, and ChatGPT Pro can summarize individual research papers fairly well. Be careful to make sure that any platform does no cherry-pick or misrepresent information.

Drafting

  • Generative AI can help with creating an outline.
    • Depending on the prompt provided, the AI can generate an outline which you can then fill in with more information. This requires that you engage with what the AI produces. 
  • Generative AI can help with distilling major topics up to a point.
    • Most platforms do not yet have a keen sense of audience, which means that they can prioritize information that may be ineffective for one audience and vice versa. Still, when you are overwhelmed with information, an AI could help you to prioritize information. 
  • Generative AI can help with summarizing and paraphrasing.
    • Platforms such as QuillBot can summarize sections of content. While this can be helpful in long writing processes, it is easily seen in writing unless you take intuitive to make the generated writing fit within your voice. Otherwise, it reads very abrupt and lifeless. 

Revision/Editing

  • Generative AI can provide feedback on grammar and sentence structure.
    • Grammarly is perhaps the most well-known with regards to sentence-level feedback. AIs can help you to recognize patterns and habits in your writing. This said, the AI is not ALWAYS right regarding its suggestions. 
    • An additional caveat: Grammarly Premium will make suggestions for full-sentence re-writes. These sentence re-writes will be found by AI detectors. 
  • Generative AI can provide a reverse outline.
    • Various platforms can provide an outline of your written document, which then can help you understand your own organization. When feeding your work into a platform, be aware of privacy concerns. That document will be within the AI's system and become "training" data.

Do not input confidential or propriety information into any generative AI platform.

 

Verify AI-generated content with reliable sources.

 

Be transparent when using generative AI and adhere to integrity guidelines.

 

Check and double-check generated content, including citations and grammar revisions.

 

Know that AI does not yet understand audience or context, which makes it particularly ineffective for particular types of writing.

 

Writing Lab consultants are also a resource to use when you feel overwhelmed or need support in your writing process.

Clemson Libraries logo.

Subject Librarians

Resources for Source Incorporation

The subject librarians can help you to find and incorporate sources into your lab reports, as well as any other writing requiring secondary sources.

Click here to find your subject librarian.
The Writing Lab
Copyright © Clemson University
The Writing Lab | 301 Cooper Library, Clemson, S.C. 29634
864-656-3606 | | Donate