In bibliometric studies, most errors happen during data collection rather than inside the software. A query that is too broad drags in thousands of irrelevant papers; an export without references rules out co-citation analysis altogether. This guide shows how to export Scopus data correctly for VOSviewer and bibliometrix, and flags what changes when you work with Web of Science.
Start with legitimate access
Scopus is a subscription database operated by Elsevier. You will normally access it through your university or institute. When exporting records for analysis, use your own institution's licensed access and follow the provider's terms of use.
Building a Boolean query with TITLE-ABS-KEY
Use Advanced search for full control. The TITLE-ABS-KEY field searches titles, abstracts and keywords, which balances recall and precision well. A typical query combines two or three concept blocks with AND, each block listing synonyms joined by OR.
TITLE-ABS-KEY ( "digital transformation" OR "digitalization" OR "digitalisation" )
AND TITLE-ABS-KEY ( "small and medium enterprise*" OR "SME*" )
AND PUBYEAR > 2009 AND PUBYEAR < 2026
- Quotation marks keep a phrase together; the asterisk truncates a word.
- Proximity operators such as W/n and PRE/n find words within n terms of each other, which helps with variable phrasing.
- Use AND NOT sparingly, since it can remove relevant papers.
- Record the exact search date. Scopus result counts change daily.
Testing and refining the query
Treat the first version of your query as a draft. Sort the results by relevance and read the titles of the first 50 records: if many are off topic, a term is too broad or ambiguous. Then sort by date or citations and repeat. Look at the Scopus refine panel for unexpected subject areas or keywords, which often reveal a second meaning of a search term. Keep every version of the query with its result count in a log file. That log takes minutes to maintain and answers the reviewer question "why these terms?" in a single paragraph.
Filtering the results
Narrow the set with the filters on the left or by adding conditions to the query itself. Filters written into the query are easier to report.
| Filter | Query syntax | Why |
|---|---|---|
| Year range | PUBYEAR > 2009 AND PUBYEAR < 2026 | Excludes the incomplete current year. |
| Document type | DOCTYPE ( ar ) OR DOCTYPE ( re ) | Keeps articles and reviews. |
| Language | LANGUAGE ( english ) | Keeps keyword analysis consistent. |
| Subject area | SUBJAREA ( busi ) OR SUBJAREA ( econ ) | Limits the field when terms are ambiguous. |
| Publication stage | PUBSTAGE ( final ) | Drops articles in press for a stable dataset. |
Log every filtering step with the number of records remaining. You will need these numbers for a PRISMA 2020 flow diagram (Page et al. 2021) if the journal asks for one.
Choosing the export format and fields
Select all results, click Export and pick a format. The three common options:
CSV
The format VOSviewer handles best, and fully supported by bibliometrix. The safest default.
BibTeX
Works well with bibliometrix and LaTeX-based reference managers.
RIS
For Zotero, EndNote or Mendeley when you need to screen records manually first.
In the field selection, tick every group: citation information (authors, title, year, source, citation count, DOI), bibliographical information (affiliations, author identifiers), abstract and keywords (abstract, author keywords, index keywords) and, above all, References. Without references there is no co-citation and no bibliographic coupling.
Before exporting, also check the total number of records once more against your log, so that any change since your last search is noticed and documented rather than discovered later.
Export limits and working in batches
Scopus caps the number of records per export, and the cap can be lower when all fields are selected. At the time of writing (November 2026), CSV exports range from a few thousand up to about 20,000 records depending on the fields chosen and your account type; check the current interface, as the provider changes these limits from time to time.
Split by year
Add PUBYEAR conditions so each batch stays under the cap, for example 2010 to 2017 and 2018 to 2025.
Export each batch
Name files consistently, such as scopus-2010-2017.csv and scopus-2018-2025.csv.
Merge and deduplicate
Combine the files in R or load them together in VOSviewer, then check for duplicates by DOI or EID. The total should match the Scopus result count.
A short note on Web of Science
In Web of Science Core Collection, the TS (Topic) field is the closest equivalent to TITLE-ABS-KEY. Export "Full Record and Cited References" as Plain text or Tab-delimited. Web of Science allows fewer records per export than Scopus (at the time of writing roughly 500 to 1,000 records with cited references; check the current interface), so you will usually export several batches by record number. Also note which indexes your institution subscribes to (SCIE, SSCI, ESCI), because subscription scope affects what you retrieve.
If you plan to combine Scopus and Web of Science, translate the query carefully rather than copying it, because field codes, truncation and proximity operators differ between the two platforms. Run both searches on the same day, export from each, and merge the records in R with bibliometrix, which harmonizes the field tags and removes duplicates. Expect a large overlap between the two databases and report how many records came from each source, how many were shared and how many remained after deduplication.
Screening and cleaning after export
Even a carefully designed query returns some records that do not belong in the dataset. How much screening you do depends on the study: a broad mapping of a large field may rely on the query alone, while a focused review usually adds a screening step.
- Title and abstract screening: import the RIS file into a reference manager or screening tool and exclude records that are clearly off topic. Record the reasons for exclusion.
- Duplicates: Scopus rarely duplicates records internally, but merged batches and merged databases do. Match on DOI first, then on normalized titles for records without a DOI.
- Missing fields: count records with no abstract, no author keywords or no references. A high share affects which analyses you can run and should be reported.
- Errata and retractions: remove errata, notes and retracted items unless they are part of your research question.
Once screening is complete, rebuild the analysis file from the original export by filtering on the identifiers you kept, rather than editing the export itself. This keeps a clean audit trail from raw data to final dataset.
Checklist before you start the analysis
Query documented
Full string, search date, database and subscription scope saved in a text file next to the data.
Fields complete
Abstracts, author and index keywords, affiliations and references present in every batch.
Counts reconciled
Records after merging and deduplication match the Scopus count minus documented exclusions.
Originals protected
Raw exports stored read-only and kept private, in line with the database licence.
Ten minutes spent on this checklist can save days of rework later, when a missing field or an unexplained difference in record counts surfaces during peer review.
Reporting data collection in your Methods
A complete data description states the database, the search date, the full search string, all filters, the record count after each step, the export format and how files were merged and deduplicated. Reviewers frequently ask for the exact query so they can rerun it.
- Put the full search string in a table or appendix.
- Justify the time span and document types.
- Acknowledge limitations: a single database, English only, indexing delays.
How Biblio VietSCI helps with the data stage
With our Scopus bibliometric analysis service, we work with you to design and test the search string, guide the export using your institution's licensed access, clean, merge and deduplicate the records, run the analyses and hand over all source files for reproducibility. The research question, interpretation and scientific conclusions stay with the author. See pricing or submit a request on the order page.
Frequently asked questions
Which format should I use to export Scopus data?
CSV is the safest choice because both VOSviewer and bibliometrix read it reliably. BibTeX also works with bibliometrix, and RIS suits screening in a reference manager.
Why can I not run a co-citation analysis?
Usually because the References field was not selected during export. Export again with all fields, including references.
How many records can I export from Scopus at once?
The limit depends on the format, the fields selected and your account, and it changes over time. Check the current interface and split the export by year if you exceed it.
Can I share the Scopus data file with my paper?
Check the Scopus terms and your institution licence first. The safe option is usually to publish the search string, search date and DOI list rather than the full records.
Should I search TITLE-ABS-KEY or TITLE only?
TITLE-ABS-KEY balances recall and precision and is the common choice. Title-only searches are more precise but tend to miss relevant papers.