Unlike general-purpose chatbots like ChatGPT, the tools listed in this guide are specifically designed to support academic research. They are powered by models trained on scholarly content and utilize a technique called Retrieval-Augmented Generation (RAG), which enables them to draw information from external sources, such as web pages or databases, to enhance the accuracy and relevance of their responses. These tools bring both benefits and risks, they are most effective when used to complement—not replace—expert search methodologies, comprehensive engagement with full texts, and rigorous critical reading practices.
Benefits
- Efficiency of literature discovery: in addition to traditional keyword searching, AI technology uses algorithms to discover similar literature via seed papers and citation chains.
- Ease of searching: AI technology allows for searching using natural language.
- Combination of free web resources and users' subscriptions: some of the listed tools allow users to import sources from referencing software like Zotero and EndNote.
Limitations
- Limitation of literature coverage: most of the listed tools use Semantic Scholar and open repositories as their data source, which retrieves open access literature and may not include sources behind paywalls.
- Most of these tools generate analysis based on abstracts but not full-text, which may miss important information and details.
- Uncertainty of search algorithms and full-text extraction technologies.
- AI does not identify good research or "bad" research, information can be generated based on poor research containing errors.
- AI systems can inherit biases present in the data they are trained on, which can affect the objectivity and accuracy of the synthesized information
- AI can summarize and organize information, but it often lacks the ability to critically analyze and synthesize complex ideas as a human researcher would.
- Due to the lack of reproducibility and transparency in semantic or vector search processes, generative AI tools can not be used solely for the Systematic Review search stage, but can be beneficial for screening and data extraction purposes.
- AI hallucinations can not be completely avoided at this stage, which conflicts with the evidence-based approach in scientific research and may lead to misleading conclusions.
- The above limitations may affect the comprehensiveness of literature discovered and accuracy of answers generated by these tools.
- There is a risk of copyright violation when using these tools' upload functionality, due to nuanced rights and permissions for different articles/journals.
- Ethical issues including plagiarism and proper attribution of sources
Suggestions
