AI-assisted Literature Review

TipLearning Objectives
  • Identify the key features and purposes of leading AI-powered literature review tools.
  • Select the appropriate tool for specific research scenarios.
  • Apply best practices for ethical and effective use of AI in the research process.

1 Background: Limitations of Traditional Literature Reviews

Traditional literature reviews, while rigorous, often face several challenges:

  • Information overload: The sheer volume of potentially relevant papers can be overwhelming, making it hard to prioritize.
  • Citation chasing limitations: Following references manually makes it hard to see large-scale relationships and influential clusters.
  • Time constraints: Researchers often lack sufficient time to systematically screen, read, and synthesize a vast literature base.
  • Summarization bottlenecks: Even with enough time, distilling and synthesizing findings from a high volume of papers is cognitively demanding and prone to errors or oversights.

2 AI tools for literature review comparison

Feature Semantic Scholar Elicit SciSpace Research Rabbit Google NotebookLM
Key Features - Paper recommendations
- Semantic search
- Displays relevant papers
- Summarizes key information
- Integrate with Zotero
- Streamline the research process
- AI writer
- Integrate with Zotero
- Discovers and visualizes relevant literature and author connections
- Integrate with Zotero
- Summarizes and organizes notes, videos, PDFs, and other documents
Price Free Free with sign-up; Pro subscription available Free basic tier; Pro subscription available Free Free, and plus (paid version)
Free Tier Limits Unlimited searches, about 50 results per page 10 top results for details and 50 results overall, up to 20–30 uploaded docs Practical ~20–50 upload, ~100 results overall ~50 results per collection, collections can grow to hundreds Up to 20–30 personal documents per notebook
Limitations - No personal uploads
- Limited customization
- Limited assistance with theoretical or non-empirical research - Hallucination in “Ask PDF” part
- Struggles with large PDFs
- No PDF uploads
- Graphs can be cluttered
- No external literature search
- Early stage beta

3 How to use it?

3.1 Scenario 1: I have a specific research question and want targeted answers

  • Recommended tools: Elicit, SciSpace, Semantic Scholar
  • Potential prompt: Risks of artificial intelligence in environmental science, specifically the content risk and ethical risk.

3.2 Scenario 2: I want to broadly discover the most influential papers in my field

  • Semantic Scholar: ideal for exploring influential papers by citation counts, influential authors, and topic clusters

3.4 Scenario 4: I have a folder of PDF articles and want to ask questions about them

  • Recommended tools:
    • SciSpace: Upload PDFs and use the AI Copilot to ask questions
    • Google NotebookLM: Upload multiple documents and explore connections, notes, and summaries

4 A list of further resources

4.1 Semantic Scholar

4.2 Elicit

4.3 SciSpace

4.4 Research Rabbit

4.5 Google NotebookLM