How to Build a Paper Search Strategy
Daniel Ha · Seoul National University PhD StudentStructure your research question using the PICO framework, then combine
Boolean search queries with AI semantic search to secure core papers. Expand
coverage through citation tracking and screen results using PRISMA criteria in
a 5-step method.
Why do you need a paper search strategy?
Without a search strategy, you risk missing important literature or wasting time sifting through thousands of results. A systematic strategy helps you search quickly and prove completeness.
Typing keywords into Google Scholar will return papers. But that alone is not enough.
| Problem | Example |
|---|---|
| Missing synonyms | Searching only "remote instruction" misses papers on "online learning" or "e-learning" |
| Too many results | Wasting time scrolling through 3,000 results one by one |
| Blocked without keywords | Beginners in a field do not know the correct academic terminology |
| Not reproducible | Without a record of how you searched, you cannot explain your process to reviewers |
A systematic search strategy solves these problems and lets you prove the rigor of your literature review to reviewers.
What is a systematic paper search method?
Structure your research question with PICO, design keywords and search queries, combine AI semantic search with keyword search, expand through citation tracking, and screen using PRISMA criteria.
Step 1: Structure your research question for search
The starting point of a search is not keywords but the research question. Use the PICO framework to decompose your question into components, and search keywords and natural language queries will follow naturally.
For example, decomposing the research question "Is AI tutoring effective for improving academic achievement in university students?" with PICO looks like this:
| Element | Question | In this study |
|---|---|---|
| P (Population) | Who is the target? | University students |
| I (Intervention/Exposure) | What is applied? | AI tutoring |
| C (Comparison) | Compared to what? | Traditional instruction |
| O (Outcome) | What is measured? | Academic achievement |
This decomposition makes extracting keywords in Step 2 easy. From P you get "university students, undergraduate, higher education"; from I you get "AI tutoring, intelligent tutoring, adaptive learning" ~ expanding synonyms for each element.
Step 2: Derive keywords and build search queries
Convert the PICO elements from Step 1 into keywords and combine them with Boolean operators to create search queries. Let us continue with the example above.
Keyword expansion
First, expand synonyms for the core keyword of each PICO element. Because different papers use different terms for the same concept, missing synonyms means missing relevant papers.
| PICO Element | Core Keyword | Synonym Expansion |
|---|---|---|
| P | University students | undergraduate, college student, higher education |
| I | AI tutoring | intelligent tutoring, adaptive learning, AI tutor |
| C | Traditional instruction | traditional instruction, face-to-face, lecture |
| O | Academic achievement | academic performance, learning outcome, GPA |
Building Boolean search queries
Combine the expanded keywords with Boolean operators to create a search query.
| Operator | Role | Usage |
|---|---|---|
| OR | Groups synonyms within the same PICO element | "AI tutoring" OR "intelligent tutoring" |
| AND | Combines different PICO elements | (P) AND (I) AND (O) |
| NOT | Excludes results outside the research scope | NOT "K-12" |
The completed search query for the example above looks like this:
("AI tutoring" OR "intelligent tutoring" OR "adaptive learning") AND ("academic performance" OR "learning outcome") AND ("higher education" OR "university" OR "undergraduate")
Step 3: Execute the search
Combining keyword search queries with AI semantic search ensures both precision and comprehensiveness.
Keyword search
Enter the Boolean search query from Step 2 into academic databases. Choose a database appropriate for your field.
| Field | Major Databases |
|---|---|
| General | Google Scholar, Scopus, Web of Science |
| Medicine/Health | PubMed, MEDLINE, Cochrane Library |
| Education | ERIC, Education Source |
| Psychology | PsycINFO, PsycArticles |
| Business/Economics | SSRN, EconLit, Business Source Complete |
| Engineering/CS | IEEE Xplore, ACM Digital Library |
AI semantic search
In the early stages when you do not know exact keywords, AI semantic search is effective. Enter your research question in natural language and it finds related papers based on meaning.
| Search Method | Input Example | Characteristics |
|---|---|---|
| Keyword search | ("AI tutoring" OR "intelligent tutoring") AND "higher education" | Requires exact terms, reproducible |
| AI semantic search | "Effect of AI-based personalized learning on university students' grades" | Natural language input, automatic synonym matching |
Nubint AI's AI Paper Search provides semantic search across 280 million academic papers. Even if a paper does not contain the keyword "AI tutoring," it finds papers about similar concepts like "intelligent learning systems" or "adaptive educational technology."
Secure 5~10 core papers first
Do not try to find every paper at once. The first goal is to secure 5~10 core papers in your field.
| Priority | Paper Type | Reason |
|---|---|---|
| 1 | Review articles | Systematic reviews and meta-analyses serve as a map of the field |
| 2 | Highly cited papers from the last 3~5 years | Show the current direction of the field |
| 3 | Papers with overlapping PICO elements | Core literature directly related to your research question |
Enter your research topic into the Literature Review agent, and the AI will analyze related papers and organize key research trends, major findings, and research gaps.
Step 4: Expand through citation tracking
Once you have secured 5~10 core papers, citation tracking is far more efficient than searching with different keyword combinations.
| Method | Direction | Description |
|---|---|---|
| Backward citation | Into the past | Follow the references cited by your core papers |
| Forward citation | Into the future | Find subsequent studies that cited your core papers |
| Snowball search | Both directions | Repeat backward and forward citation tracking to expand related papers. See Wohlin (2014) |
Nubint AI's Research Flow Explorer agent automatically analyzes the citation network of a specific paper and shows key prior works and follow-up studies.
Step 5: Screen and organize
Filter the papers gathered through your search step by step following the PRISMA guidelines. After removing duplicates, screen by title and abstract first, then read the full text of remaining papers to determine final inclusion.
| Stage | Task | Example |
|---|---|---|
| Identification | Total search results combined | 1,240 |
| Duplicate removal | Remove duplicates across DBs | → 890 |
| Title/abstract screening | First-round selection by relevance | → 120 |
| Full-text review | Apply inclusion/exclusion criteria | → 45 |
| Final inclusion | Papers used in analysis | → 32 |
Organize selected papers immediately. If you put it off, you will lose track of sources. Save them in reference management tools like Zotero or Mendeley, and add notes on inclusion/exclusion rationale, key findings, and methodology for each paper ~ this makes writing your literature review much easier later. Saving papers to Nubint AI's Paper Library also allows agents to reference them during analysis.
What paper search sites are available?
Google Scholar and PubMed are free, Scopus and Web of Science are accessible through university libraries, and Nubint AI supports AI semantic search.
| Site | Features | Free? |
|---|---|---|
| Google Scholar | The most versatile academic search engine. Provides citation tracking and full-text links | Free |
| PubMed | The largest database for medicine and life sciences. Operated by the U.S. National Library of Medicine | Free |
| Scopus | International journal article search and citation analysis. Operated by Elsevier | Subscription |
| Web of Science | Search for SCI/SSCI-indexed papers and citation analysis | Subscription |
| Nubint AI | AI semantic search across 280 million papers. Natural language input, field-agnostic | Subscription |
If you are affiliated with a university, you can access paid databases like Scopus and Web of Science for free through your library portal. Combine multiple sites depending on your search purpose, but starting with AI semantic search lets you quickly find core papers even in the early stages when you do not know the right keywords.
How is AI paper search different from keyword search?
AI semantic search lets you start quickly with natural language, while keyword search excels at reproducible systematic searches. Combining both is the most effective approach.
| Comparison | AI Semantic Search | Keyword (Boolean) Search |
|---|---|---|
| Input method | Natural language questions | Boolean operators + exact terms |
| Strength | Automatic synonym matching, useful for initial exploration | Reproducible, essential for systematic reviews |
| Weakness | Hard to reproduce the search query | Requires knowing exact keywords to start |
| Best stage | Early research, exploring core papers | Comprehensive search after finalizing strategy |
Start with AI search, then supplement with keyword search. In the early stages of research, you often do not know the exact terminology. By first securing core papers through AI semantic search, you can identify the precise keywords used in your field and then build Boolean search queries for systematic expansion.
Nubint AI's AI Paper Search provides semantic search across 280 million academic papers. Without searching multiple databases separately, a single natural language question can find related papers across fields.
Summary
The key to paper searching is not what keywords to enter, but what question you need the papers to answer. Secure core papers quickly with AI semantic search, expand the network through citation tracking, and supplement with keyword searches. Recording your search process lets you prove the rigor of your literature review.