Semantic Scholar & Connected Papers for Citation Mapping

Once you have found a few good papers, the next problem is finding the rest of the relevant literature without spending three weeks digging through reference lists by hand. This is where citation-mapping tools become genuinely transformative. Semantic Scholar, Connected Papers, ResearchRabbit, and scite let you explore the network of papers around a seed paper — and AI features inside them surface what is actually important.

This lesson teaches the workflow that turns one good paper into a complete literature map in under an hour.

What You'll Learn

• How citation networks work and why they matter for finding the best sources
• How to use Semantic Scholar and Connected Papers as a beginner
• The "seed paper" workflow for building a literature map
• How to use scite to check whether a paper has been supported or contradicted by later work

Why Citation Mapping Is the Best Hidden Skill

Every academic paper cites earlier papers (its references) and may be cited by later papers (the "cited by" list). Together these form a network. Papers in the same network tend to be on the same topic. Papers cited by many other papers in a network are usually foundational. Papers that cite a lot of newer work are usually reviews or syntheses.

Before citation-mapping tools existed, exploring this network meant manually clicking through reference lists one paper at a time. It took days. Now it takes minutes — but the skill of reading the map well still takes practice.

Semantic Scholar (semanticscholar.org)

Free, no login required for basic use. Indexes over 200 million papers across all disciplines. When you find a paper page, you get:

• The abstract. As a baseline.
• TLDR. A one-sentence AI-generated summary of the paper's main point. Useful for quick triage. Treat as a hint, not a substitute for reading.
• References. Every paper the authors cited, with links.
• Citations. Every paper that has cited this one, with links and dates. You can sort and filter.
• Related Papers. AI-recommended similar papers based on shared content and citation overlap.

The killer feature for a student is the Citations tab. If you find a foundational paper in a field, you can sort its citations by recency to see what current research is doing with that idea. If you find an old paper that "everyone cites," you can find the more recent paper that synthesized it.

You can also create a free account to save papers to a personal library, follow authors, and get alerts when new papers cite something you care about.

Connected Papers (connectedpapers.com)

Free for a limited number of graphs per month; subscribe for unlimited. You enter one paper (your "seed paper") and Connected Papers generates a visual graph: each circle is a paper, lines connect papers that share many citations, larger circles indicate more cited papers, darker shades indicate more recent papers.

What you do with the graph:

• Identify foundational papers. The biggest circles in the graph are usually the most influential papers in this neighborhood of the literature.
• Find recent work. Darker shaded circles are recent — useful for understanding the current state of the conversation.
• Spot clusters. The graph often forms visual clusters; each cluster usually represents a sub-topic or a research tradition.

Connected Papers also offers "Prior" and "Derivative" views — papers that came before your seed paper (intellectual ancestors) and papers that built on it (descendants). These are gold for tracing how an idea evolved.

The Seed Paper Workflow

This is the workflow most professional researchers use. Try it on a current course topic.

Step 1: Find one good paper on your topic. Use Elicit, Perplexity, or a paper your professor mentioned. This is your seed paper. Quality matters here — pick a well-cited, peer-reviewed paper that is directly on topic.

Step 2: Open it in Semantic Scholar. Read the TLDR and abstract. Look at the references and the citations.

Step 3: Generate a Connected Papers graph. Paste your seed paper's title or DOI into Connected Papers. Wait for the graph.

Step 4: Identify three types of nearby papers.

• A foundational paper (large circle, older): an earlier classic that your seed paper built on.
• A review paper (look for "review" or "meta-analysis" in the title): summarizes the field.
• A recent paper (darkest circle): the latest word on the topic.

Step 5: Read those three. Use the AI tools from earlier lessons to summarize them. Now you have a much richer map of the literature than you did 20 minutes ago.

Step 6: Pick a new seed paper. Often the foundational or recent paper makes a better seed than your original. Repeat the process. After two or three rounds, you have a strong sense of the literature.

This is what literature reviews are made of.

scite (scite.ai)

A tool with a unique angle: for any paper, it tells you how many later papers support, contrast, or mention its claims. Most citations are neutral mentions, but the ones that explicitly contrast or support a paper are extremely valuable.

If you cite a paper in your own work, scite lets you check whether later research has reinforced or contradicted its finding. This is a strong signal of how the claim has held up over time. Free tier offers limited features; many universities have an institutional subscription — check with your library.

A paper that is mostly "mentioned" but rarely "supported" is being cited politely but not actually used. A paper that has many "contrast" citations is being argued against in later work — sometimes a sign of a foundational but contested claim, sometimes a sign that it has been superseded.

ResearchRabbit (researchrabbit.ai)

Free, similar to Connected Papers but with a smoother workflow. You can create collections of seed papers and watch the related-paper graph evolve as you add more. It is particularly nice for long-running projects like a thesis or dissertation, because the graph keeps updating as you save more papers.

You do not need to use every tool. Pick one citation-mapping tool (Connected Papers or ResearchRabbit) and one citation-impact tool (Semantic Scholar plus scite if your school provides it). That covers most undergraduate and master's needs.

Worked Example

Suppose your seed paper is a well-known study on social media and political polarization. You generate a Connected Papers graph and see three clusters: one around "filter bubbles" and algorithms, one around "selective exposure" in political psychology, one around "cross-cutting exposure" in communication research.

Now you understand something important: there are three different research traditions converging on this question, with different assumptions and methods. A weak paper would cite three papers from one cluster and call it a day. A strong paper acknowledges all three traditions, even if it focuses on one.

You could not have seen this in a single afternoon five years ago. With these tools, you can see it in 20 minutes — but you still have to read the papers to understand what each cluster is actually arguing.

A Quick Exercise

Pick a paper from a current course — any paper you have already read, even an assigned reading. Search for it on Semantic Scholar. Read the TLDR. Generate a Connected Papers graph. Identify one foundational paper, one review, and one paper from the past 12 months that cites your seed paper.

You have just built the skeleton of a literature map. Save the four titles in your notes.

Key Takeaways

• Citation networks reveal the structure of a literature in a way no AI summary can replicate.
• Use Semantic Scholar for fast paper discovery and to find what cites or is cited by a paper.
• Use Connected Papers or ResearchRabbit to visualize the neighborhood around a seed paper and find foundational, review, and recent works.
• Use scite (if available) to check whether a paper's claims have been supported or contradicted by later research.
• The workflow is seed → references and citations → graph → read three new papers → pick a new seed → repeat.