AI as Pre-Search Scaffolding:
Integrating GenAI into Information Literacy at Connecticut College
This spring, I partnered with four STEM faculty to explore a question at the heart of modern research: How can generative AI help students find relevant sources efficiently, without replacing the essential research process? Our answer: Use AI as pre-search scaffolding, then bring students into library databases with targeted terms, clear strategies, and the skills to evaluate what they find.
Our Framework: “AI suggests, databases verify.”
We designed each session around four steps:
Define the Target: Clarify the assignment’s learning goals.
AI for Ideation Only: Use GenAI to brainstorm keywords, concepts, and search strategies.
Database Verification: Move immediately into library databases to execute and refine searches.
Reflection Check: Ask, “What did AI miss or get wrong?” and document findings.
Image created with Adobe Express
Course Highlights:
Computer Science Research Seminar (400 level)
The students in this course had already participated in an information literacy instruction session in other courses. The instructor wanted the students to experiment with using GenAI for topic refinement and keyword expansion.
Suggested prompts:
I am a Computer Science undergraduate working on a senior research project about [INSERT TOPIC]. Help me narrow this down to 3 specific, researchable questions suitable for an academic paper.
Act as an academic librarian. I am researching [INSERT NARROWED TOPIC]. Please generate a list of 15 keywords and phrases related to this topic. Include synonyms, alternative spellings, and related technical terms. Group them by concept.
Using the keywords we just generated, write a Boolean search string for the database Scopus. Use parentheses to group synonyms with OR, and connect main concepts with AND.
The students transferred these suggestions to Scopus and Science Direct for more target searching. Students also used ChatGPT and Perplexity within Boodlebox to search for literature on their topic. They then checked the results against academic databases to verify that the article actually exists.
Biology (200 level)
Students needed to choose “some” additive or drug or growth condition to test on yeast cells. Ideally, they would choose something that had not been tested on yeast before, so they needed to research the effects of “something” on other organisms or cells to help them formulate their hypotheses. Because this approach might require several iterations of trial and error to identify the “something” that would be tested, the students were encouraged to prompt Google Gemini to help streamline the process.
Suggested prompt:
I am an undergraduate student designing a biology experiment to test how a specific variable affects yeast cell growth. My variable is [Insert Variable, e.g., Caffeine].
Provide:
1. The scientific/chemical name for this variable.
2. A list of 5 related keywords or synonyms I should use in Scopus.
3. Three specific biological mechanisms this variable is known to affect in
other organisms (e.g., metabolism, membrane stability).
This enabled the students to quickly verify the novelty of their variable (or not) by effectively searching for its absence in Saccharomyces cerevisiae literature.
Data, Information and Society (200 level)
This is the foundational course for the Data, Information and Society Pathway. For this session, we mainly focused on identifying the potentiality for bias in datasets sourced from Google Gemini. Specifically the learning goals included:
Identify key questions to ask when starting a data search.
Use library databases, AI tools (Google Gemini) and open data sources to locate relevant datasets.
Evaluate datasets for credibility, objectivity, and completeness using a systematic approach.
Critically assess AI-generated data recommendations for bias, accuracy, and completeness.
Recognize how algorithmic bias in AI tools can lead to biased research outcomes.
In practice students were asked to predict what sources Google Gemini might prefer when asked for data sets on a particular topic. They then searched for data in an academic database (ICPSR or Data Planet) and Google Gemini and compared the results. This lead to discussion about verification and evaluation of data, retrieval bias, and why it matters in research.
Biology (100 level)
For this course I had the opportunity to meet with students for two sessions. The first session focused solely on traditional research methods. The second session introduced the use of Google Gemini to help develop the research strategy. The instructor expressed frustration with past students not being able to find literature that met the assignment criteria. In their words, “Most students [in this class] don’t know disciplinary vocabulary yet. How do they get to relevant literature if they don’t know what/how to find it?” With this in mind we developed a scaffolded series of prompts to help guide the students through their search.
Suggested Prompts:
Prompt 1 - Generate your Search Strategy
I’m an undergraduate student working on a biology research assignment. I am required to find the following:
a) Three primary research articles about body size or wing size in Drosophilia.
b) One research article about insecticide resistance and any trade-off it has with another morphological or physiological trait. You can look specifically for DDT as the pesticide, but may have better luck finding an article with the generic “insecticide” or “pesticide” terms.
What specific search terms, synonyms, scientific terminology, and Boolean operators should I use when searching in biology databases like Scopus, Science Direct, or Science in Context? Please organize these as a search strategy I can actually use, including alternative phrasings.
Prompt 2 - Understand What Makes a “Good” Article
Using the search strategy you just provided, can you help me identify what types of peer-reviewed research articles I should be looking for? Specifically:
1) What would distinguish a good research article from a review article for this assignment?
2) What experimental methods might I expect to see in quality research on this topic?
3) What are key indicators that an article presents original data versus summarizing others’ work?
Prompt 3 - : Evaluate Your Found Articles
I found these articles in [database name]: [student pastes 2-3 article titles and abstracts]. Can you help me evaluate whether these meet my assignment criteria? I need to assess:
1) Are these peer-reviewed research articles with original experimental data?
2) Do they directly address my research question?
3) What are the key findings and methods?
4) Are there any limitations or concerns I should note?
5) Would these be appropriate sources for an undergraduate evolution paper?
Prompt 4 - Expand Your Search
Based on the articles I’ve found, can you help me identify:
1) What are 2-3 additional search terms or related concepts I should explore to find more sources?
2) Are there specific researchers or labs whose work I should follow up on?
3) What related evolutionary concepts might broaden my understanding of this topic?
Using these prompts the student used Scopus and Science Direct for searching and were able to efficiently discover results that were aligned with the assignment criteria. The students were then able to complete their annotated bibliographies during class time which allowed for discussion and feedback with the instructor as they were designing their potential lab experiments through informed research.
An opportunity for teaching about academic integrity “in the moment” presented itself during this session. As I was demonstrating the prompting method in Google Gemini the tool offered this:
Because the deliverable for this assignment was an annotated bibliography, accepting this offer would clearly overstep the boundaries of acceptable GenAI use for this assignment. This led to a robust discussion of recognizing when and how GenAI might cross an academic integrity line.
Assessing the Process:
Our students quickly embraced the “AI suggests, databases verify” framework, finding it a powerful tool for navigating the complexities of academic research. Students shared that they found the additional GenAI layer to be extremely helpful to their research strategy and would use it again in the future if allowed for an assignment.
In follow up conversations with the faculty they were pleased with the outcome, particularly in regard to increased efficiency in the identification of relevant resources.
Looking forward to future collaborations I would like to have the opportunity to talk more about AI literacy itself and allow more time for students to practice using GenAI tools with guidance. I would also encourage instructors to add GenAI use guidance/boundaries to the assignment itself, rather than recycling an assignment and retrofitting it in.
Declaration of AI use in my process:
Formalizing the acknowledgment of AI tools in academic workflows is crucial for transparency and academic integrity. This ensures clarity for both instructors and students.
I used Gemini 3 Pro and ChatGPT5 (Boodlebox) to develop prompts for the lessons based on the learning objectives. I used Gemini 3 (Boodlebox) to synthesize the content in the four lesson plans and used that information to create an outline for a Table Talk presentation. From that presentation I wrote a draft for this article and then used ChatGPT 4.1 (Boodlebox) to improve formatting for Substack.