There are many AI-based tools in development that purport to help you with “research.” A non-exhaustive list would include: Semantic Scholar, Elicit, Research Rabbit, Perplexity, Consensus, and Scite. They vary in the types of sources they index and the format of their search results, but they commonly attempt to summarize and distill large and complex amounts of information for you.
On the surface they seem tremendously helpful, but can they replace JSTOR or your favorite library database? Not yet. Here are a few tools to look at more closely – they all draw on Semantic Scholar’s corpus of sources, but they present content in very different ways.
Semantic Scholar: In existence for a few years already, Semantic Scholar partners with publishers in STEM and social sciences to access scholarly content. When you type in a search, the results will look similar to Google Scholar, with the exception of the TLDR article summary. Click “Expand” at the end of the summary and you can read the article’s abstract. If you click on the article title you can follow the citation network for the paper: its references, citations, and “related papers.” The drawback appears to be its search algorithm, which does not always understand your search string or question. Nevertheless, you can potentially discover different sources than if you use Google Scholar alone.
Elicit: Elicit borrows Semantic Scholar’s corpus of literature and puts an entirely different user interface and searching algorithm on top of it. It synthesizes the “Top” 4 papers on your topic and also provides summaries of the papers themselves. Look closely at the paragraph synthesis, however, and compare its conclusions with the article summaries: they may be at odds. Elicit also admits that it cannot always distinguish a good study from a bad one. The natural language searching feature is very good though, and it can be useful to students exploring topics.
Research Rabbit: Research Rabbit also borrows content from Semantic Scholar. It is primarily a citation explorer, rather than a search engine, and works best if you “feed” it a few papers first. For this task, it can sync with Zotero and upload collections or folders. From there, explore references, citations, and “similar work.” Confirmation bias is a risk, of course, but nevertheless and it can be revealing to “go down a rabbit hole,” as it were, as you uncover the scholarly communication thread.
Overall, literature research stands to greatly benefit from AI-based tools, especially when you are navigating citation networks, uncovering buried sources, and exploring research topics. But don’t remove your JSTOR bookmark yet.
