Skip to Main Content

AI Tools for Academic Literature Research

This guide includes examples of available AI-powered tools to assist with scholarly literature searches and reference management plus ethical considerations and warnings.

Open laptop on desk surrounded by stacks of paper journals. Drawn images appear to be emerging from screen in addition to a yellow lightbulb.

Laptop image generated using Adobe Firefly Model 3 and prompt: Open laptop on a desk surrounded by journal articles and other research references. Images of journal articles, book chapters, formulas, and diagrams emerge from the screen and appear to diffuse into the air around it. Also emerging from the laptop is a lightbulb as a metaphor for a significant idea or revelation. Use a surrealist style and blue and purple tones.

Six Ways Diagram

Integrating AI Tools Into the Research Process

Hexagon diagram showing six ways AI might be used in Academic Research

AI for Scholarly Literature Research diagram by Kimberly S. Grotewold is licensed under CC BY-NC 4.0 

6 Ways that AI Can Support Your Academic Literature Research Process

  1. Identify general research topic ideas
  2. Refine research topic ideas and questions (relevant subtopics, topic intersections, etc.)
  3. Plan a search strategy (identify useful databases to search, formulate keyword and phrase searches)
  4. Search for sources and reveal relationships among sources and authors*
  5. Manage sources and notes
  6. Assist with writing mechanics 

Visit the pages of this guide to learn more.

*Many AI tools such as Perplexity, ChatGPT, Gemini, Microsoft Copilot, Research Rabbit, Connected Papers, etc., are limited in the content they can retrieve and reference when compared to what is available through the library's subscription-based databases.

Prompt Frameworks

Role, Task, Format 

This approach is a basic prompt structure useful for beginners.

  • Role: Start by assigning a role to the AI. This helps the model understand the expected knowledge and style. For example, "You are an expert academic researcher."
  • Task: Clearly define the task you want the AI to perform. Use action verbs to indicate the specific action (e.g., "write", "summarize", "analyze"). For example, "Write an annotated bibliography on the topic of college students' uses of generative AI.
  • Format: Be specific about the output structure, presentation, and file type (Note: the output may be limited to certain file types.) For example, "Use APA 7th Edition style and format. Include 10 research-based journal articles. Each annotation must include 2-3 sentences summarizing the article, 1-2 sentences on the major findings, and 1-2 sentences on the article's implications for the field. The output should be a text file. For each reference, include a DOI or a web link for accessing the full-text of the article."

(Anthropic, 2025; Grotewold, 2025)

 

Content Based on:

Anthropic. (2025, October 6). AI prompt frameworks for academia [Generative AI chat]. Claude Sonnet 4.5 with Web search, Research, and Extended thinking. https://claude.ai/public/artifacts/15cb4c58-de09-4d09-9a04-1726870e9b2e

Grotewold, K. S. (2025, June 9-12). Navigating Ai in academic literature reviews: Aligning efficiency with ethics [Conference session]. Texas A&M University System Chancellor's Conference on Academic Technology, Bryan, TX. United States. 

CLEAR 

This approach emphasizes prompter engagement, thinking, and analysis.

  • Concise: Include the essential information in the prompt and exclude extra details.
  • Logical: Structure information or instructions within prompts in step-by-step order.
  • Explicit: Use specific instructions about content, format, and scope of the desired output.
  • Adaptive: Expect to re-prompt the model(s) or system(s) being used, providing additional context or key terms.
  • Reflective: Continuously evaluate AI-generated outputs and human prompts, considering accuracy, bias, and relevance.
    (Anthropic, 2025)
     

Content Based on: 

Anthropic. (2025, October 6). AI prompt frameworks for academia [Generative AI chat]. Claude Sonnet 4.5 with Web search, Research, and Extended thinking. https://claude.ai/public/artifacts/15cb4c58-de09-4d09-9a04-1726870e9b2e

Lo, L. S. (2023). The art and science of prompt Engineering: A new literacy in the information age. Internet Reference Services Quarterly, (ahead-of-print), 1-8. https://doi.org/10.1080/10875301.2023.2227621

 

Prompt Framework Shared by Greg Brockman from OpenAI

This prompt structure is similar to RFT and RACE but includes the unique feature of reminding the AI model/system to check specific details that tend to be prone to mistakes. 

  1. Goal: Define what you want as clearly as possible
  2. Response Format: Specify how the response should be structured (list, paragraphs, json, etc.)
  3. Warnings: Point out details the AI should double-check
  4. Context: Provide background information 

Adapted from Tulane University Libraries, AI and Academic Research Guide: Prompt Engineering, which is licensed under CC BY-NC 4.0

CRISPE

The experiment step offers the prompter output options to choose from, forcing them to evaluate the output and exercise some control over which (if any) output is used. 

  • Capacity/Role: Define the function or capacity in which the AI functions, for example, botany researcher.
  • Insight: Provide significant background information and context.
  • Statement: Provide a clear specification of the task to be accomplished.
  • Personality: Use tone/personality instructions or settings to guide the style of the response.
  • Experiment: Ask for an experiment by requesting multiple options or strategies. 
     

Content Based on: 

Anthropic. (2025, October 6). AI prompt frameworks for academia [Generative AI chat]. Claude Sonnet 4.5 with Web search, Research, and Extended thinking. https://claude.ai/public/artifacts/15cb4c58-de09-4d09-9a04-1726870e9b2e

RACE 

Including specific contextual information, such as file examples, can assist in the generation of useful outputs.

  • Role: Define the role the AI should assume, including any specialized expertise. For example, "Act as an experienced cognitive psychology researcher with expertise in mixed methods study design and analyzing text as data."
  • Action: Use precise verbs to specify the action the AI model should use before generating output.
  • Context: Provide additional background information, situational details, audience characteristics, and sample model documents or files.
  • Execute/Expectations: Tell the AI what to produce with clear specifications about output, format, and structure, or how many examples/choices to offer.

(Anthropic, 2025)


Content Based on:

Anthropic. (2025, October 6). AI prompt frameworks for academia [Generative AI chat]. Claude Sonnet 4.5 with Web search, Research, and Extended thinking. https://claude.ai/public/artifacts/15cb4c58-de09-4d09-9a04-1726870e9b2e