Large Language Models are technology that can help automate several tasks associated with research (in addition to many other applications). They are trained on text and can help with reasoning tasks, answering questions, and searching along the citation chain for scientific literature. They are changing rapidly, but currently they show promise for research but certainly don't replace the need for critical thinking and subject expertise.
| AI Tool | Best Research Use | Cost | Features at a Glance |
|---|---|---|---|
|
|
Developing Research Topics; Background Research |
Free Pro: Starting at $20/month |
Uses: Have students ask the chat to ask them questions about their topics or give an overview of a topic. Use a specific prompt and ask Chat GPT to provide citations. Considerations: Students must actually find and read the sources and check citations to make sure they are real. Tools like Chat GPT may "hallucinate" and provide citations that are not actual sources. It is a starting point, not an ending point of research. |
| Perplexity |
Developing Research Topics; Background Research |
Free Pro: $20/month |
Uses: Ask Perplexity a research question and it will answer your question using web-based sources. Considerations: Students should evaluate how accurate these responses are, and look closely at the sources returned to make sure these are the best version to answer. Ethical concerns with copyright and how Perplexity is gaining access to sources. |
| Research Rabbit |
Visualizing scholarly conversation; Finding scholarly articles |
Free |
Uses: Start with a seed article and then visualize linkages between scholarly citations and authors. You can look at a timeline of articles published about a topic, or find related authors or citations. Considerations: Are these citation mappings comprehensive? It's unclear. Research Rabbit hasn't published much about their methodology. |
| Paper Pilot |
Visualizing scholarly conversation; Finding scholarly articles |
Free: 5 graphs/month Pro: starting at $85/month |
Uses: Ask a question and get a research-backed answer using scholarly sources. List of references on the side. Considerations: Not comprehensive, papers may be out of date. Students will have to evaluate sources generated (and the answer) to see if there are errors or gaps. |
| Elicit |
Developing a Literature Review; Extracting data for Systematic Reviews |
Free Pro: $12/month |
Uses: Ask a question and get a research-backed answer linking to top papers. There will also be a grid pulling information from papers such as brief summary, sample size and location, study type, and much more. It is responsive and customizable. Considerations: As always, students and faculty must evaluate the accuracy of information pulled from sources. Elicit is a powerful tool built to look at scholarly literature, but can "hallucinate" data or other information pulled from sources. |
| NotebookLM |
Read and Understand Scholarly Articles |
Free |
Uses: Upload PDFs of your research and have NotebookLM summarize or create a podcast. Ask questions and get answers based on the papers you uploaded. Concerns about hallucinations are much smaller because you're feeding it the papers. Considerations: Like all AI tools, it can be vague and not particularly in-depth in how it summarizes the work. Must still read the scholarly material to evaluate inaccuracies. |
| Scholarcy | Read and Understand Scholarly Articles |
Free: 3 summaries/day Pro: $10/month |
Uses: Load a paper into Scholarcy to extract a summary or highlight important parts of the paper to assist with interpreting scientific literature. Create flash cards that outline how the paper relates to other papers in the field. Considerations: As always must evaluate the accuracy of summaries and analysis. |
| Julius | Data analysis and data visualization |
Free: 15 messages per month Pro version with account upgrade |
Uses: Use natural language queries to interact with data. Load data using spreadsheets or datasets and create graphs, forecasting models, and more. Considerations: May generate misleading or incorrect visualizations. May prioritize clarity over depth. AI models can inherit bias from training data or user inputs. |
For more comprehensive listing of AI Research Tools, see Aaron Tay's Musings about Librarianship and Best AI Tools to Write A Literature Review.
AI is a machine that makes mistakes. It may be an okay background source (similar to how we talk about Wikipedia), but if you're relying on it for research, you should follow the citations and actually read the sources it references.
Chatbots are trained on publicly-available data, but that data comes from a predominantly white, male, English-speaking, American perspective. And, since AI doesn't have a mind of it's own to filter out biases, it will often simply amplify the biases found in the source material.
AI uses a huge amount of power and water. By some estimates, "a request made through ChatGPT...consumes 10 times the electricity of a Google Search" and globally in 2024, consumes 6x the water of Denmark (UN Environment Program).
Chat-based tools can quickly and easily summarize content, making it tempting to copy-paste text. Students may have trouble differentiating between appropriate uses of AI to help with scholarship, and uses that cross the line into plagiarism.
