What each tool does for biomedical literature mapping

Connected Papers and BioSkepsis Research Landscape both produce interactive network graphs of research papers. The similarity ends at the surface.

Connected Papers takes a single seed paper, analyses roughly 50,000 papers from the Semantic Scholar corpus, and selects the few dozen most connected by co-citation and bibliographic coupling. The output is a force-directed graph where node proximity reflects similarity. Users can view prior works (highly cited ancestors) and derivative works (papers that cite the graph). There is no AI layer: no synthesis, no full-text reading, no biological reasoning.

BioSkepsis Research Landscape starts from a natural-language research question. The system semantically searches 40 million+ curated biomedical papers, retrieves up to 1,000 per session, and maps them as an interactive network. But the graph is only the starting point. BioSkepsis reads the full text of selected papers, including methods sections and supplementary data, then generates citation-grounded narrative syntheses, identifies emerging trends, classifies each paper's structural role in the network, and proposes testable hypotheses.

Connection types: co-citation vs biological multi-layer networks

Connected Papers relies on two metrics. Co-citation measures how often two papers are cited together by a third paper. Bibliographic coupling measures how many references two papers share. Both are metadata-level signals; neither requires reading a word of the actual research.

BioSkepsis identifies five types of connections between papers in the Research Landscape. Shared genes: papers studying the same gene products. Pathway connections: research investigating related biological pathways or processes. Methodological links: studies using similar experimental approaches. Citation networks: direct citation relationships. Semantic similarity: papers with related conceptual content identified through deep semantic analysis of full text.

The difference matters for biology. Two papers can study the same gene via completely different methods and never cite each other. Connected Papers may never surface that link. BioSkepsis, reading the full text and mapping Gene Ontology terms, connects them directly.

Structural paper classification in the biomedical network

Connected Papers distinguishes between "prior works" and "derivative works" relative to the graph. Prior works are older, highly cited papers that the cluster references. Derivative works are newer papers that cite papers in the cluster. This is useful for situating a field temporally, but it assigns no structural role to individual papers within the network.

BioSkepsis classifies every paper in the Research Landscape into one of four structural roles based on its position in the citation network and its biological content. Foundational Papers are seminal works that anchor entire research programmes. Hub Papers are highly connected nodes that many other papers reference or build upon. Bridge Papers connect separate research clusters, often representing interdisciplinary or translational work. Novel Leads are under-recognised papers with high biological relevance but low citation counts, representing potential opportunities the field has not yet fully incorporated.

For researchers conducting systematic reviews or planning new experiments, knowing which paper is a bridge between two sub-fields, or which novel lead deserves closer attention, is qualitatively different from knowing which papers are simply older or newer.

AI synthesis and biological reasoning: the core divide

Connected Papers does not use AI for analysis. It does not read papers. It does not generate summaries, syntheses, or interpretations. Multiple independent reviews confirm this: it is a pure discovery and visualisation tool. Researchers must read and synthesise the discovered papers themselves, or export them to a separate tool.

BioSkepsis layers AI-driven analysis on top of the Research Landscape graph. Once papers are mapped, the system reads their full text and produces a narrative synthesis of the field's structure, key debates, and knowledge gaps. It detects emerging trends by identifying research frontiers where more than half of high-impact publications are from the last three years. It generates testable hypotheses grounded in the synthesised literature. It builds mechanistic link tables showing molecular and biological connections across papers.

Every AI-generated claim in BioSkepsis is traceable to a specific passage in a specific paper. If the evidence is insufficient, the system says so. This citation grounding is the structural difference: Connected Papers shows you a map; BioSkepsis reads the territory.

Corpus scope and biomedical domain specificity

Connected Papers draws from the Semantic Scholar corpus, which indexes over 200 million papers across all academic disciplines. This breadth is a strength for interdisciplinary researchers but means the tool has no domain-specific retrieval logic. A graph built from a molecular biology paper may include tangentially related work from computer science or social sciences if the citation patterns overlap.

BioSkepsis indexes 40 million+ curated biomedical papers from 1931 to present, updated weekly. Retrieval is biology-native: it uses Gene Ontology annotations, MeSH descriptors, gene names, and domain-specific keywords. When the Research Landscape expands a graph, it draws from Semantic Scholar's broader corpus but filters through biomedical relevance scoring. The result is a network that reflects biological relationships, not just citation coincidence.

For researchers working in biomedicine, pharmacology, agricultural science, or ecology, this domain specificity means fewer irrelevant nodes and denser, more informative clusters.

Entry point: seed paper vs research question

Connected Papers requires a seed paper. You must already have one relevant paper in hand before you can build a graph. This works well for researchers who are deep in a field and want to explore the neighbourhood of a known paper. It works less well for researchers entering a new field, scoping a review, or investigating a question they cannot yet pin to a single paper.

BioSkepsis accepts a natural-language research question. Type a question about TNF-alpha signalling in rheumatoid arthritis, and the system retrieves, maps, and analyses the relevant literature without requiring a seed paper. This question-first approach is closer to how researchers actually think: the question comes before the bibliography.

Connected Papers does support keyword search to find a seed paper, and multi-origin graphs allow iterative expansion from multiple seeds. But the fundamental unit remains the paper, not the question.

Pricing, access, and biomedical research plans

Connected Papers offers a free tier limited to five graphs per month. Paid plans range from approximately $3/month (Academic, annual billing) to $15/month (Business). All plans include the same features; the distinction is usage volume and commercial licensing.

BioSkepsis offers a free Basic plan that includes full Research Landscape access, AI synthesis, and all analytical features, limited by monthly token budgets (roughly five research sessions). Paid plans (Plus at EUR 8/month, Pro at EUR 35/month, Team at EUR 60/month per seat) expand monthly capacity, context windows, and team collaboration features. All plans include the Research Landscape, hypothesis generation, trend detection, and full-text analysis.

The pricing comparison is not apples-to-apples. Connected Papers charges for graph volume. BioSkepsis charges for the depth of AI reasoning applied to a much larger analytical surface. A BioSkepsis session includes network mapping, full-text reading, synthesis, and structured analysis; a Connected Papers session produces a static visual graph.

Who should use which biomedical research landscape tool

Frequently asked questions