Image
News
Writing with AI Chatbots
Article by Hannah Jacobs
Academic writing can be hard. It takes considerable thought and creativity. When we’re having trouble, it can be helpful to have some assistance structuring our ideas, simplifying our syntax, or aligning our work with a CFP or NOFO.
What this post is not: a guide to get chatbots to write for you.
What this post is: an opportunity to try some ways to push past writer’s block, see your work in a different way, and decide how you want to move forward.
Duke’s ChatGPT Edu, Copilot, or DukeGPT are all tools available to support AI-assisted writing. With your latest draft open in your favorite word processor, try some of the following prompts:
Brainstorm a paper title:
Act as a [your subject area] scholar writing a paper for [your journal or conference’s name] about [briefly describe your topic, making sure to include important keywords]. Create a title for the paper. Here are two examples:
- Example A
- Example B
Create 5 more possible titles for this paper.
For example:
Act as a digital humanities scholar writing a paper for Digital Humanities Quarterly about using minimal computing principles to create an archive of oral histories recorded with developers about the evolution of their work with genAI. Here are two examples:
- The GenAI Evolution Project: Building an Oral History Archive with Minimal Computing
- A Minimal Computing Approach to Oral History Archives: The GenAI Evolution Project
Create 5 more possible titles for this paper.
Prompt to create paper titles & responses in ChatGPT Edu using GPT-5.
Structure an outline:
Act as a [your subject area] scholar writing a paper about [briefly describe your topic, making sure to include important keywords]. You need to include the following themes:
- Subtopic A
- Subtopic B
- Subtopic C
- Subtopic D
Create a detailed outline that includes possible subtopics.
For example:
Act as a digital humanities scholar writing a paper about using minimal computing principles to create an archive of oral histories recorded with developers about the evolution of their work with genAI. You need to include the following themes:
- Principles of Minimal Computing
- The structure of the project & project team
- Why the team chose minimal computing
- Challenges encountered when creating the archive
Create a detailed outline that includes possible subtopics.
Image
Prompt to create an outline & part of response in DukeGPT using GPT-4.1.
Edit a conference or journal abstract from an outline:
Act as a [your subject area] scholar submitting an abstract to [relevant publication or conference name] about [briefly describe your topic, making sure to include important keywords]. Here is the CFP: [link to CFP]. Here is the draft abstract:
[Your abstract here.]
Suggest revisions to this abstract for clarity and grammar, keeping in mind the themes highlighted in the CFP and the [number of words]-word limit.
For example:
Act as a digital humanities scholar submitting an abstract about using minimal computing principles to create an archive of oral histories recorded with developers about the evolution of their work with genAI to the Digital Humanities 2025 conference. Here is the CFP: https://dh2025.adho.org/call-for-proposals/. Here is the draft abstract:
[Abstract here.]
Suggest revisions to this abstract for clarity and grammar, keeping in mind the themes highlighted in the CFP and the 250-word limit.
Image
Part of prompt to revise an abstract & response in Microsoft Co-Pilot using GPT-4.
Remember: You are the expert! Only you know the nuances of your topic. A chatbot should not be trusted for accurate descriptions or analyses. You may need to adjust your prompt to get something closer to what you’re aiming for. Then it’s your turn to take what the chatbot offers and refine it further into a piece that fits your goal.
Some important considerations:
- Environmental impact: All digital technologies require natural resources including rare minerals and electrical power. Generative AI currently uses considerable resources. Learn more what it takes, calculate your energy usage, and consider using prompt engineering to limit the number of prompts you need.
- Intellectual property: Some chatbots (like ChatGPT) use human prompts to train future models, and sharing your unpublished intellectual property could lead to your work showing up in a chatbot response to someone else. Be sure to check a chatbot’s privacy policy: do they train models on user prompts? If so, what implications might that have for how you do or do not use them? Duke’s ChatGPT Edu, Copilot, and DukeGPT do not train models on user prompts.
- Publisher, Funder, & Organization Policies: Funders, publishers, and professional organizations are beginning to set policies around AI use in academic submissions. Some funders, like the NIH, are prohibiting the use of AI when writing proposals. Some publishers, like the Association for Computing Machinery, are embracing AI use and clarifying how it may be used and cited. Others, like Nature, allow for limited use in text while prohibiting use for image and video. Here is a list of some publishers and their AI policies. If you can’t find a policy, ask before you draft. Even if your organization or publisher doesn’t have a policy, it’s always best practice to document and cite your use of AI. Note that AI chatbots cannot be considered co-authors because they cannot assume responsibility for a written work.
Further Resources for Designing Prompts
“Prompt Patterns.” Generative AI at Vanderbilt. Vanderbilt University.
“Prompt Frameworks.” Prompt Engineering for Academic Research. New York Institute of Technology Library.
“GenAI Prompt Guide.” Stanford University IT.
“Prompt Literacy in Academics.” Generative Artificial Intelligence. University of Michigan.