Connected Papers is an AI-driven literature graph analysis platform designed for researchers, applied scientists, and academic explorers. Its core positioning is to solve the pain point of traditional literature retrieval—"seeing the trees but not the forest"—by building dynamic visual graphs based on citation relationships and co-citation networks, helping users quickly identify core papers, milestone research, and recent advances in a field, thereby greatly improving the efficiency of literature review and survey writing.

The platform was launched in 2020 by a team in Tel Aviv, Israel, leveraging advanced graph algorithms (such as cosine similarity, bibliographic coupling, and co-citation clustering) to conduct deep mining of large academic databases like Semantic Scholar and OpenAlex. Its technical foundation integrates natural language processing and complex network analysis, enabling real-time generation of topological relationships between any two papers and adaptive display of the field's evolutionary trajectory. In the AI academic search arena, Connected Papers has become a benchmark product distinct from traditional keyword search, thanks to its unique "visual exploration" paradigm, and is widely acclaimed by researchers in interdisciplinary fields such as computer science, biomedicine, and physics.

• Literature Relationship Graph - Input any paper (DOI, URL, or title) to automatically generate an undirected network graph containing 50-100 highly related papers. Node size represents citation count, color represents publication year, and edge thickness represents similarity strength—allowing you to grasp the field's structure at a glance.
• Prior Works / Derivative Works - Algorithms automatically identify the paper's foundational works (Prior Works) and subsequent derivative research (Derivative Works), helping users quickly build a cognitive chain from classics to frontiers.
• Similar Papers List - A list of papers sorted by similarity is displayed alongside the graph; each entry can be expanded to show abstract, citation count, year, and authors, with one-click access to the original paper or an "Expand" option for further exploration.
• Keyword Highlighting and Filtering - Supports highlighting and filtering the graph by keywords (e.g., "Transformer," "CRISPR") to precisely pinpoint literatures related to specific technical approaches.
• Export and Sharing - One-click export of the graph as PNG or JSON, with support for generating shareable links (including temporary editing permissions), ideal for presenting the full literature landscape in paper meetings or project reports.
• Seamless Integration with Academic Ecosystem - Integrates major databases such as Semantic Scholar, arXiv, PubMed, and ACM, and provides quick reference import and citation integration with tools like Zotero and Mendeley.

• Registration and Access - Visit the official website (connectedpapers.com), click "Sign Up" in the upper right corner, and register for free using a Google account or email. Free users can use basic graph features without payment and without needing to download or install any client.
• Input a Seed Paper - In the search box on the homepage, paste the paper's DOI (e.g., 10.1038/s41586-020-2649-2), an arXiv link, or directly type title keywords. The system will automatically match the most relevant paper; click the "Build Graph" button to start generation.
• Explore and Manipulate the Graph - Once the graph is generated, you can drag nodes to adjust layout, hover over nodes to view abstracts and citation details. Use the left control panel to switch between "Prior Works" or "Derivative Works" and filter specific year ranges via the timeline at the bottom. Click the "Expand" button to add more related papers until you achieve a comprehensive view.
• Export and Collaboration - When the graph reaches the desired structure, click "Export" and choose PNG or JSON format to save. For online sharing, click "Share" to generate a unique link (with read or edit permissions) and send it to collaborators via email or chat tools, enabling multi-user simultaneous editing of the graph.

Example 1: Quickly Surveying a New Field
Prompt: Enter DOI: 10.48550/arXiv.2112.10752 (a paper about GPT).
Expected Output: The platform generates a graph centered on GPT, automatically displaying upstream and downstream key papers such as Transformer, BERT, GPT-2/3. Through "Prior Works," users can trace the origin of the attention mechanism (Attention Is All You Need), and via "Derivative Works," they can see papers related to ChatGPT applications—building a three-dimensional understanding of the evolution of language models within 10 minutes.

Example 2: Interdisciplinary Literature Exploration
Prompt: Enter the paper title "Deep learning in drug discovery".
Expected Output: The graph will contain representative papers in directions such as molecule generation, protein structure prediction, and virtual screening. By using keyword filtering and inputting "graph neural network," all papers related to graph neural networks are immediately highlighted, allowing quick selection of the latest research for citation, followed by direct export as PNG for inclusion in a proposal report.

Example 3: Tracking Recent Advances in a Field
Prompt: Enter a familiar paper such as "YOLOv5" and then click "Derivative Works".
Expected Output: The system lists all papers from the last two years that cite YOLOv5 and are highly similar to it (e.g., YOLOv7, YOLOv8, YOLO-NAS), sorted by publication year and citation count, helping researchers keep up with technological iteration trends.

• Ultimate Visualization Efficiency - Compared to traditional search result lists, the graphical display lets users quickly identify core papers, key branches, and research gaps without reading every title and abstract, reducing literature review time by more than 50%.
• Intelligent Citation Network Algorithms - Uses advanced graph embedding and co-citation analysis techniques that go beyond direct citations to uncover deep topic similarities, discovering cross-field associations that traditional keyword search can hardly capture.
• Zero Learning Curve - No need to learn any search syntax; a single seed paper generates the whole field graph, making it extremely friendly to novice researchers.
• Freemium Model - Core features are completely free; paid subscriptions only unlock advanced features like batch export and expanded graph nodes, making it very cost-effective for individual academic users.
• Continuously Updated Database - The underlying data comes from Semantic Scholar's daily updated index, ensuring that newly published papers (especially preprints) are captured by the graph within days, maintaining cutting-edge relevance.

Similar competitors include Semantic Scholar's Research Feeds, Litmaps, and SciSpace (formerly Typeset).

• Comparison with Litmaps - Litmaps also offers visual literature maps, but Connected Papers emphasizes "similarity" rather than mere citation chains, enabling faster discovery of non-directly related papers in a field, with a cleaner interface and lower learning cost.
• Comparison with Semantic Scholar - Semantic Scholar excels at search and recommendations, but its "Related Papers" list is linear and lacks a global view. Connected Papers combines visual exploration with semantic similarity, making it better suited for deep, survey-type reviews.
• Comparison with SciSpace - SciSpace focuses on paper interpretation and Q&A, while Connected Papers specializes in relationship network construction. They are complementary, but Connected Papers excels in algorithmic granularity for citation networks (e.g., separate analysis of "derivative works").

Connected Papers adopts a Freemium model. Currently (as of 2026), the following tiers are available:

• Free - Permanently free. Allows creation of up to 10 graphs, each containing approximately 50 core nodes, with basic export (PNG with watermark) and shareable links that expire after 7 days. Suitable for personal initial exploration.
• Pro - Priced at approximately $5/month (annual billing) or $8/month (monthly billing). Includes unlimited graph creation, expansion to 100 nodes, watermark-free PNG export, full JSON export, priority processing queue, and shareable links valid for 1 year. Ideal for graduate students and lecturers who frequently conduct literature reviews.
• Team - Billed per seat at approximately $12/month per seat (minimum 5 seats). Supports shared workspaces, centralized invoicing, SSO login (SAML), and an admin control panel. Suitable for collaborative literature research in university research groups or corporate R&D centers.
• API and Custom Solutions - For libraries, publishers, or research platforms, API access (usage-based pricing) and on-premise deployment options are available; specific pricing requires contacting sales. Academic institutions typically receive educational discounts on Team plans.

• Academic Researchers and PhD Students - Those who need to quickly write paper introductions, literature reviews, or identify research gaps. Connected Papers helps them build field maps, especially useful for interdisciplinary researchers looking to understand a new field quickly.
• Applied Scientists and Industrial R&D Personnel - In corporate research environments such as AI, biotech, and materials science, this tool helps track frontier algorithms and competitor technology trajectories, supporting technology selection.
• University Lecturers and Graduate Advisors - Used for course preparation (e.g., "Introduction to Machine Learning"), guiding graduate students through thesis proposals, or building shared literature knowledge graphs for research groups.
• Scientific Editors and Journal Reviewers - Quickly verify the relevance and completeness of references in manuscripts, improving review efficiency and quality.

• Academic Proposal and Review Writing - After identifying a research direction, generate a field graph from a seed paper to quickly collect 80% of the relevant core literature, organize technological evolution by timeline, and build the skeleton of a review.
• Grant and Funding Applications - Use the graph export as a high-resolution image for the "Research Status and Trends" section, visually demonstrating the complex relationships within the research field and enhancing the proposal's persuasiveness.
• Industry Technology Monitoring and Competitive Analysis - R&D teams input competitor patents or core technical papers and use "Derived Works" to monitor subsequent improvement directions, identifying potential collaboration or avoidance opportunities.
• Graduate Student Academic Training - Advisors guide students to use the tool to organize the classic-mainstream-latest three-layer literature structure, helping them avoid blind searching and develop critical reading habits.
• Interdisciplinary Knowledge Transfer - A biomedical researcher inputs a paper on deep learning applied to image analysis; the graph displays bridge nodes between related medical and computer science literature, promoting cross-team collaboration.