What Is Connected Papers and How to Use It for Research

TL;DR: Connected Papers is a website that turns one research paper into a map of the few dozen papers most similar to it. You choose a starting paper, and it shows you the older papers your topic is built on, the newest follow-ups, and how they all group together. That makes it a quick way to get your bearings in a topic you're new to. To use it well, open the papers it finds, check where each one was published, and look for the question nobody has tested yet. If finding those papers makes you want to run a study of your own but you haven't built an AI model yet, Veritas AI's Combination program takes you from AI basics in small mentor-led classes to an independent project you design one-on-one with a mentor.

If you are a high schooler with a research question, you might have found one great paper and then gotten stuck. Its reference list runs for pages, and it's hard to tell which sources really matter. Connected Papers helps with exactly that moment: you give it one paper, and it draws a map of the research around it. You don't need an account or any coding to try it.

Say you're curious whether AI can spot fake product reviews, the kind you scroll past when you shop online. One well-known study on that question is enough to start. From that single paper, Connected Papers can show you which studies started the field, which ones are newest, and where the open questions sit. You'll follow that same fake-reviews example all the way through the steps below.

What is Connected Papers?

Connected Papers is a website that builds a visual map of research papers related to one paper you choose. That starting paper is called the origin paper. Every paper on the map shows up as a bubble, and papers on similar ideas sit close together. You get a picture of a whole research area at once, instead of one long list of search results.

Connected Papers began as a weekend side project between friends who wanted better tools for reading research. The team launched it in 2020, and by 2022, more than two million researchers had used it in a single year. Its papers come from the Semantic Scholar database, which holds more than 200 million papers across many fields. That's why one tool can map topics as different as fake-review detection, climate models, and cancer research.

How does Connected Papers work?

Connected Papers works by checking which papers share the same sources and the same readers. For each graph, it scans about 50,000 papers and keeps the few dozen with the strongest links to your origin paper. It measures those links in two ways. The first is co-citation, which means two papers often get cited together by later studies. The second is bibliographic coupling, which means two papers cite many of the same sources. When two papers overlap a lot in either way, they're very likely about the same topic.

Connected Papers then arranges the map so similar papers pull together and less similar ones drift apart. Two papers can sit right next to each other even if neither one cites the other. In the fake-reviews example, one study might say "opinion spam" while another says "fake reviews." If they lean on the same sources, they still end up side by side. That's how Connected Papers finds papers a keyword search would miss.

How to read a Connected Papers graph

You read a Connected Papers graph through four visual clues: bubble size, bubble color, distance, and lines. Each one answers a different question about the papers on your map, as library guides from HKUST and other universities explain. Here's what each part means and how you can use it.

Part of the graph	What it shows	How to use it
Bubble size	How many times the paper has been cited	Big bubbles are papers the field leans on, so read those early
Bubble color	When the paper came out: darker is newer, lighter is older	Check the darkest bubbles to see where the topic is heading now
Distance between bubbles	How similar two papers are	A tight cluster usually means one sub-topic, like fake hotel reviews
Lines between bubbles	A strong similarity link between two papers	Follow the strongest lines to find the papers most like the one you're reading
Clicking a bubble	The paper's details, plus its shortest path back to your origin paper	Use it to see how a far-off paper connects to your starting point

How to use Connected Papers for research: 6 steps

You can go from one paper to a full reading list in six steps. Here's how each step plays out with the fake-reviews question.

1. Pick an origin paper. Your origin paper should sit right at the center of your question. If you don't have one yet, search your topic on Google Scholar or Semantic Scholar and choose a study that's been cited a lot. For fake reviews, a strong pick is Finding Deceptive Opinion Spam by Any Stretch of the Imagination, a 2011 study from Cornell. Its team built a model that spotted fake hotel reviews with nearly 90% accuracy, while human readers did little better than guessing. Don't worry if a paper like this feels hard to read yet, since you only need its title to start. Still looking for a question of your own? These AI research ideas for high school students can spark one.

2. Build the graph and scan the list view. Paste the title into the search bar on Connected Papers, click the paper, and your map appears. Then open the list view to see each paper with its year and citation count. You can filter the list by keyword, by year, or to open-access papers that anyone can read online.

3. Open Prior Works to find the foundations. Prior Works lists the older papers that many papers on your graph cite. These are usually the studies your whole topic grew out of. Reading two or three of them first makes the newer papers much easier to follow.

4. Open Derivative Works to find the newest research. Derivative Works lists papers that cite many papers on your graph. They tend to be recent studies or review articles that sum up the topic. Sort them by year to see what researchers are testing right now.

5. Add a second origin paper to sharpen the map. If your graph feels too broad, add another paper that matches your question closely. Connected Papers then builds a new graph of papers similar to both. You could pair the hotel-review study with a paper on product reviews from online stores.

6. Save your papers and look for the gap. Export your list as a BibTeX file, a format citation tools read, so your references are ready when you write. Then ask what none of the papers on your map has tried yet. Maybe most studies test hotel reviews, and you wonder whether the same methods work on reviews for phone apps.

Turning that kind of gap into a question you can actually test is where a mentor makes the biggest difference. If you're ready to build an original AI project around your question, Veritas AI's AI Fellowship with Publication & Showcase gives you 15 one-on-one sessions with a mentor from a top university or AI lab. You also get 2 sessions with a publication expert, so sending your work to a journal becomes a real possibility.

What are the limitations of Connected Papers?

Connected Papers is great at showing you where to look, but it can't do the reading or judging for you. Knowing its five main limits helps you plan around each one.

• The basic plan caps your graphs. Connected Papers limits the basic plan to 5 graphs a month. Choose your origin papers carefully so each graph counts.

• Brand-new papers can be hard to spot. The map depends on citations, so recent studies and very niche topics may not have enough links to show up strongly. Search Semantic Scholar by date to catch the newest work.

• It doesn't read papers for you. You see titles, years, and abstracts, but no summaries of what each study found. Open the papers that matter and read their results yourself.

• Similar doesn't mean reliable. A paper can sit close to your origin paper and still be a preprint that no experts have reviewed yet. Check where each paper was published before you cite it.

• Your map won't update on its own. Connected Papers doesn't alert you when new papers on your topic appear. Rebuild your graph a few weeks into your project to see what's changed.

Connected Papers vs ResearchRabbit vs Litmaps: which is best for research?

No single tool wins here, since each one is built for a different stage of a project. Connected Papers gives the fastest first look, ResearchRabbit grows with you over a long project, and Litmaps shows how a topic changed over time. Here's how they compare on the things that matter most when you're choosing.

	Connected Papers	ResearchRabbit	Litmaps
How it maps papers	A similarity map built around one origin paper	Recommendations that grow from a collection of papers you save	A citation map that places papers by year and citation count
Best for	A quick snapshot of a topic that's new to you	Collecting and organizing papers across a long project	Seeing how a topic built up year by year
Basic plan limits	5 graphs a month	Up to 50 seed papers and unlimited searches	Up to 20 search inputs and 100 articles per map
Alerts	No built-in alerts	Research integrity alerts on its paid plan	Monthly alerts on the basic plan, weekly or daily on Pro

Pick Connected Papers if you're starting a brand-new topic and want one clear map fast. Pick ResearchRabbit if you'll keep adding papers for months and want to share your collection with a teammate or mentor. Pick Litmaps if you care most about how a topic grew over time and want updates when new papers come out. You can also mix them, using Connected Papers for your first look and ResearchRabbit or Litmaps as the project grows. ResearchRabbit also shows up among these AI research tools for high school students if you want more options to try.

FAQs

1. Is Connected Papers an AI tool?

Not in the way a chatbot is, since Connected Papers doesn't write text or answer questions. It measures how closely papers are linked through their citations and lays them out as a map. Its paper data comes from Semantic Scholar, which uses AI models built in-house to process and label each paper. So you can think of Connected Papers as a smart map built on top of AI-processed data.

2. How much does Connected Papers cost?

Connected Papers has a basic plan and two paid plans. The basic plan includes every feature, including multi-origin graphs, but caps you at 5 graphs a month. The Academic plan covers academic researchers, nonprofits, and personal use, and the Business plan covers commercial use. Both paid plans unlock unlimited graphs. The Academic plan costs $6 a month, billed as $72 once a year.

3. Is Connected Papers good for beginner researchers?

Yes, it's one of the easiest research tools to start with, since one paper title is enough to build a whole map. You don't need to know a field's big names or special search terms first. The harder part for a beginner usually comes next: knowing what to do with all those papers. If you're new to both research and AI, Veritas AI's AI Scholars program teaches Python and AI in live classes at a 5:1 student-to-mentor ratio, with no coding experience needed. You finish by building a real AI research project with 3 to 4 peers.

4. Can Connected Papers help with a literature review?

Yes, it's a great place to start one. A literature review sums up what's already been published on your question, and it often forms the introduction of a research paper. Papers that sit in the same cluster on your map often share a theme, so each cluster can become one paragraph of your review. You'll still need to read each paper and explain how they connect in your own words. These steps on how to write an AI research paper in high school show where your review fits in the full paper.

Where to learn more about Connected Papers

These pages are the best places to go deeper once you've built your first graph.

• Connected Papers About page: How the team builds each graph, in their own words.

• Connected Papers plans announcement: What the basic and paid plans include, plus how multi-origin graphs work.

• About Semantic Scholar: Where Connected Papers gets its paper data, and how that database uses AI.

• HKUST Library guide to citation mapping tools: A university library's walkthrough of Connected Papers, with tips on reading the graph.