Literature reviews are foundational to academic research, enabling researchers to contextualise their work within existing knowledge. Historically, researchers relied on manual searches and citation chaining, a systematic but labour-intensive method of tracing references backward and forward. The advent of digital databases in the 1990s, such as Web of Science and Scopus, introduced more scalable solutions, yet these still required significant expertise in query formulation. The integration of AI and network visualisation in the 2010s marked a paradigm shift, with tools like VOSviewer enabling researchers to map citation networks (van Eck & Waltman 2010). Today, generative AI tools build on these foundations by combining citation-based mapping with semantic search capabilities, offering intuitive interfaces and dynamic visualisations to support both novice and experienced researchers.
ResearchRabbit (Dixon 2025), a free tool launched in 2021, is designed for exhaustive literature exploration. It allows researchers to create collections of papers, which serve as "seeds" for generating citation-based graphs. These graphs connect papers based on references and citations, with nodes representing papers and edges indicating citation relationships. A key strength of ResearchRabbit is its speed, capable of identifying thousands of papers. Its integration with Zotero facilitates seamless reference management, and features like author search and timeline views enable researchers to explore a field’s evolution and key contributors. However, its interface can be overwhelming, often presenting thousands of results without clear prioritisation, which may challenge users seeking focused outcomes. For example, a researcher studying ecological modelling might receive suggestions for broadly cited but irrelevant papers, such as those referencing the R programming language.
Litmaps (Kisley 2025), a subscription-based tool, distinguishes itself through its intuitive 2D graph layout, where the x-axis represents publication year and the y-axis indicates citation count. This design allows researchers to quickly identify recent and highly cited papers. Litmaps supports multiple seed papers, enabling more comprehensive exploration than single-paper-based tools. Its monitoring feature notifies users of new relevant papers, and co-authorship search identifies works by authors within a collection, enhancing discovery of niche literature. While praised for usability, Litmaps’ free version is limited to two graphs, and its premium plan (£10/month) may deter budget-conscious researchers. Additionally, its batch processing of 10 papers at a time can slow large-scale reviews.
Connected Papers (Behera et al. 2023), introduced in 2020, focuses on simplicity and speed, generating graphs from a single seed paper using Semantic Scholar’s metadata. It employs bibliographic coupling and co-citation analysis to cluster thematically similar papers, with node size reflecting citation count and colour indicating publication year. Features like "Prior Works" and "Derivative Works" help identify foundational and emerging studies, respectively. While its user-friendly interface is ideal for quick snapshots, the limitation to one seed paper restricts its ability to map broader fields comprehensively. The free version caps usage at five graphs per month, with a premium plan at £3/month. For researchers in interdisciplinary fields, such as digital humanities, Connected Papers may miss non-journal sources due to its reliance on Semantic Scholar.
While citation-based tools dominate, semantic search tools like Elicit (Byun 2025), SciSpace (Jain et al. 2024), and Consensus (Apata et al. 2025) leverage AI to analyse paper abstracts or full texts, enabling plain-text queries. These tools can be useful for early-stage research or when precise keywords are unknown. However, their reliance on AI can lead to less reliable results, as semantic similarity may not reflect true relevance. For instance, ecological studies on plants and animals may share similar terminology but belong to distinct subfields, leading to irrelevant suggestions. Recent advancements, such as Litmaps’ semantic search feature introduced in 2024, attempt to bridge this gap by combining citation and semantic approaches, though its performance remains slower and less precise.
References:
1. Apata, Olukayode E., Oi-Man Kwok, and Yuan-Hsuan Lee. 2025. The Use of Generative Artificial Intelligence (AI) in Academic Research: A Review of the Consensus App. Cureus 17(7). ^ Back
2. Behera, Prashanta Kumar; Jain, Sanmati Jinendran; & Kumar, Ashok. 2023. Visual Exploration of Literature Using Connected Papers: A Practical Approach. Issues in Science and Technology Librarianship, (104). DOI: https://doi.org/10.29173/istl2760 ^ Back
3. Byun, Jungwon. 2025. Introducing Elicit Systematic Review. Published 20 February 2025 on the Elicit blog. Available at: https://blog.elicit.com/systematic-review/ ^ Back
4. Dixon, Digl. 2025. Using ResearchRabbit to Speed Up Literature Review. Published 18 June 2025. https://www.researchrabbit.ai/articles/using-researchrabbit-to-speed-up-literature-review ^ Back
5. Kisley, Marina. 2025. Litmaps Database: Coverage, Open Access, and Article Indexing. Available at: https://docs.litmaps.com/en/articles/7212085-litmaps-database-coverage-open-access-and-article-indexing ^ Back
6. van Eck, Nees Jan, and Ludo Waltman. 2010. Software Survey: VOSviewer, a Computer Program for Bibliometric Mapping. Scientometrics 84, 523–538. https://doi.org/10.1007/s11192-009-0146-3 ^ Back
