First Time User's Guide to Connected Papers for Research

Connected Papers turns the messy task of finding relevant research into a visual, similarity-based map. Instead of reading dozens of papers just to guess what matters, it builds a graph from a starting point—searched by title, DOI, or URL—and then clusters related work so researchers can quickly see where a field’s ideas concentrate and how different papers connect.

The interface settles into a three-panel layout: a list of retrieved papers, a central network graph, and a side panel showing the selected paper’s abstract. The seed paper (the origin) appears in purple, and the graph’s structure is the key to understanding what to do next. The connections are not a traditional citation tree. Papers are positioned and grouped by similarity—derived from citation patterns across many works—so clusters represent areas with shared “understandings” and research themes rather than direct “paper A cites paper B” relationships.

Three visual cues guide exploration. First, line thickness reflects how strongly papers are connected to the seed and to each other, producing tight clusters for closely related topics while leaving “lonely” papers isolated when they don’t match the field’s dominant similarity neighborhoods. Second, node size indicates the number of citations: heavily cited papers appear larger, signaling broad attention in the scientific community. Third, color encodes publication time: lighter nodes are older, while darker nodes are more recent. The practical takeaway is to look for papers that are both large and dark—highly cited and relatively new—because they often combine influence with current relevance.

Connected Papers also supports workflow decisions. In list view, results can be sorted by similarity to the origin, year, citations, or references, and the set can be downloaded as a BibTeX (bib) file for import into reference managers. Two sections are especially useful: “prior works” highlights earlier foundational papers that many heavily cited works build on, while “derivative works” surfaces newer papers that cite multiple items in the graph, pointing to where the research is heading.

A less intuitive feature involves blue-highlighted papers. Blue indicates a paper is cited by the “selective derivative works” in the network, meaning it sits on a direct citation path from highly connected later work. Even though the layout is similarity-based, this blue cue helps identify items that are both central to the cluster and directly referenced by other important papers.

The tool’s right panel makes it easy to open abstracts and jump to external sources such as Semantic Scholar, publisher pages, Google Scholar, and PubMed. It also offers filters for keyword, open access/PDF availability, and year. Finally, researchers can refine the map by creating new graphs from a selected paper (“open graph”) or adding multiple origins (“add origin”) to thicken and expand the network; origins can be removed to recalculate the graph. The result is a fast, repeatable method for building a reading list and aligning new research with established and emerging work.