AI Tools for Academic Literature Research

Elicit

• Elicit pulls from the Semantic Scholar articles database, so it has access to over 200 million academic papers.
• One of Elicit's main features is its capacity to generate a literature review matrix in table format based on a user's query.
• The free version used to have more features, but some of these features have shifted to requiring a paid subscription (e.g. the ability to download the literature matrix generated now requires a paid plan).
• The free version retrieves resources relevant to the user's query, presents a summary of the "top 4 papers," and automatically generates a table listing the papers and an abstract summary. Users can add two additional columns to the table without an upcharge. Elicit suggests possible columns to add but users can also enter their own prompt.
• In the free version, options to change the sorting of results and to filter the papers according to whether a pdf is available, publication date, the quality quartile of the journal, and study type. Searching abstracts for specific keywords is also a free feature.
• Selecting papers and searching their citation trails is also free. The cited papers are added to the list of papers in the table.
• Another currently free feature is the capacity to abstract concepts with definitions/explanations from the papers.
• The free version allows the user to share their output page (view only). Click the image to see the output shown.
• Adding more columns to the literature matrix table including one "high accuracy column," generating a summary of the "top 8 papers," doing more extensive "data extraction," and exporting the results as CSV, RSI, and BIB files requires a paid subscription.
• In the free level plan, the data extraction feature can be used with 10 papers (uploaded as pdfs) per month.
• As of January 2025, the Elicit Plus Plan for individual researchers costs $12/month, which a Pro Plan for systematic reviews is $49/month.

View larger image

• The image shows a search for articles related to the query: "What is the difference between explainable AI (XAI) and interpretable AI? Which term is applicable to making AI processes more transparent and understandable to non-technical audiences?" 
• Elicit generated a one-paragraph summary of the top 4 papers with in-text citations that link to articles in the Notebook (the collection of articles retrieved). In this case the searcher added two columns to the basic table. The additional columns shown are Methodology and Main findings. The sidebar with column suggestions is shown.

For more information, see the Elicit Help Center. You can also join their Slack channel.

ResearchRabbit

• ResearchRabbit is a free, powerful, AI-enabled search tool that will recommend additional sources and create visualizations of the research landscape for the user's topic. Note: It does not generate summaries or other text outputs.
• It uses OpenAlex and Semantic Scholar as data sources and claims to be the largest academic database or resources other than Google Scholar. It uses search algorithms developed by Semantic Scholar and the National Institutes of Health (NIH).
• To begin, the user starts a collection by naming it. Then the user must upload at least one source document. Documents can also be uploaded automatically from a reference manager (e.g. Zotero, Mendeley).
• Selecting one or multiple source documents offers the user options for finding additional, related resources: Similar Work, All References, All Citations, [Works by] These Authors, [Works by] Suggested Authors, or Linked Work. Click the image to see the map online.
• Similar work is likely to retrieve the largest number of related sources because it pulls from references and citation data plus some "additional magic."
• The related papers are displayed in a column to the right of the function selections with abstracts and links to full-text if available.
• ResearchRabbit offers two main visualization options: Connections and Timeline.
• In the Connections View: Circles are generally authors/works and lines between circles represent co-author relationships. Circles are clustered according to topics.
• In the Timeline View, circles are still authors/works but they are arranged according to date of publication. Co-authorship lines are not visible in this view.
• From the Similar Works, the user can then expand a collection by viewing other works by the identified authors (These Authors), works by Suggested Authors, and Linked Content...
• So far, the functionality has only been described based on using a single source as a starting point. It is easy to become overwhelmed in ResearchRabbit, especially when starting with a collection of multiple sources and building increasingly complex visualizations and lists of sources.
• Lists of papers included in collections can be downloaded in CSV, BibTex, and RIS formats. Visualizations are downloadable as png files. Users can attach comments to sources as a way of including their own notes in the process.
• ResearchRabbit may not work well for very current topics lacking enough research to analyze for connections.

View larger image

• The image offers a view of the process beginning with one article, "A History of Instructional Media, Instructional Design, and Theories," and selecting Similar Work to create a Network Graph visualization by first author of the similar works data set. 
• An oval encircles the next step choices of exploring These Authors (from the Similar Works group), Suggested Authors (algorithm-determined), or Linked Content. 
• Alternately, a user could go back through and make different choices at each step (shown by a panel) and/or add more papers to the collection to generate additional options and paths.

For more information, see the ResearchRabbit home page, or view their Welcome to ResearchRabbit video.

Connected Papers

• Connected Papers is connected to the Semantic Scholar paper database so it has access to around 200 million papers.
• No log-in is required to try the features. Without a log-in, users can generate 2 graphs per month.
• Users can search by keywords, paper title, DOI, or other identifiers
• Upon selecting a paper as a starting point, Connected Papers builds a graph based on that seed paper. The graph shows papers that are similar based on overlapping citations and references (not just the papers cited in the seed paper).
• Papers that are most similar are clustered together and have stronger connecting lines.
• Circle (Node) size indicates the number of times a paper has been cited.
• Different viewing options include: Prior Works, Derivative Works, and List View tables. Additional filters include by keyword, PDF availability, and publication date.
• Free users with a login can generate up to 5 graphs per month. Paid plans that allow unlimited graphs start at $6/month.

View larger image

• The gold arrows point out the original paper used to create the graph.
• The purple rectangle encloses the menu of options for changing the graph view and for filtering papers to be included in the selection.

For more information, see the Connected Papers About page.

Litmaps

• Litmaps provides access to over 270 million research articles. 
• The search function of Litmaps is built on open access metadata from Crossref, Semantic Scholar, and OpenAlex.
• Litmaps is available as a website and a mobile app.
• Users can search by keyword, author, DOI, Pubmed ID, or arXiv ID.
• Litmaps produces visualizations showing relationships among articles. These relationships are determined using the following:
  • Shared citations and references
  • Common authors
  • Similar text--this function involves AI-powered semantic analysis
• A free use level with no log-in required provides basic search of up to 20 inputs and 2 Litmaps per month with 100 articles per map.
• Setting up a log-in offers options for setting up collections and iterating on prior collections and maps.
• Users can start with a search and create a basic, auto-generated map.
• More advanced uses include adding keyword tags (represented by colors) and producing maps by adding selected articles to a map and building more intentionally.
• A Pro Educational license is available for $10/month and has advanced search capabilities, plus unlimited inputs, articles, and Litmaps. Team and Enterprise-level accounts are also available. There is even a Teach with Litmaps program.
• Litmaps allows for importing multiple articles at once and also syncs with Zotero.

View larger image

• The image shows the seed article by Weng (2024) marked in gold rectangle above the Explore Related Articles column and in a gold oval at the bottom right corner of the visualization map.
• The map shows twenty related articles arranged according to publication date, number of citations and relatedness to the seed article.
• Visualizations can be downloaded and additional views are available.
• Paid users can view more than 20 articles at a time and can add articles to their maps and lists.

For more information, see the LitMaps Features page. You can also sign-up for their Substack.

Consensus

• Consensus is built on the Semantic Scholar content database with access to approximately 200 million papers. It uses keyword search and approximate nearest neighbor (ANN) algorithm-powered vector search across titles and abstracts to retrieve results.
• Consensus offers multiple functions at the free level. A free account provides up to 10 Pro Analyses per month.
• A user can enter a keyword search, question, or other prompt. 
• In response, Consensus provides a response or answer that draws from the "top 10 most relevant papers."
• Citations to the papers are included in the response. Although, the Pro Analysis summarizes 10 papers to create a response, it may not include in-text citations for all 10 papers.
• In addition to the initial response to the user's prompt, recommended follow-up or related questions are offered. Then the 10 papers that were used to draft the response are listed.
• Highly cited papers and paper types like preprint and systematic reviews are marked.
• If the full-text of a paper in the list is available, the user will see an option to "Ask this Paper." This feature allows the user to query the individual paper for a summary, key takeaways, etc.

View larger image

• Scrolling further, three related questions are suggested.
• Next, the 10 articles used to create the generated response are listed with labels noting publication date, article type, number of citations, etc.

View larger image

• Finally, the "Ask this paper" feature is indicated with a purple arrow.

For more information, see the Consensus Help Center.

Scite_ (Scite.ai)

• Without a login, users can try 2 free prompts.
• Setting up a login begins a 7-day free trial. Personal accounts are available for $12/month. Discount prices for students and academics are available "if you recommend Scite to your institution." Institutions can subscribe to Enterprise accounts.
• Scite has access to 187 million publications.
• A user can enter a question or other content into a prompt box. Users can also enter a specific article title or DOI.
• A multi-paragraph response is generated that includes linked in-text citations.
• In addition to the AI-generated response, a side panel opens that offers two views: References and Search Strategy.
  • References lists the sources used to create the response with citations and links.
  • Search Strategy presents a list of searches used including links to the searches. The user can manually edit the searches to change the outputs.
• Another option is the Table View, which allows the user to create an exportable, multi-column, literature review matrix with sources and extracted content.
• The user can adjust settings before entering a query using the gear icon or by clicking on Settings.
  • "Always use references" can be selected to ensure that citations are incorporated into the generated response paragraphs.
  • The user can also select the number of references that should be consulted before generating the response.
• With a paid account, the user can enter follow-up questions, revisit the menu, etc.
• Additional paid account features include a user dashboard, browser extension, and a Zotero plug-in.

View larger image

For more information, see the Scite_ landing page, especially the Product dropdown menu. Additionally, the Scite development team published a journal article in 2021, explaining how they built the tool.