Is Google Scholar an acceptable database for a systematic review?

Not as a primary source. Its results are not reproducible, the interface caps retrieval, and the search syntax is limited. It is useful as a supplementary check and for grey literature, and it should be reported as such, with the date and the number of results screened.

Do I have to search in languages other than English?

You do not have to, but you must report the restriction and consider its effect. In fields with substantial non-English literature, notably Chinese-language exercise and rehabilitation trials, an English-only restriction can bias the evidence base. If you restrict, say so in the limitations and explain why.

How do I report a search that was updated during the review?

Report both dates and what the update returned. A sentence such as "The search was run on 3 February 2026 and updated on 18 August 2026; the update identified 412 additional records, of which 3 met the inclusion criteria and were added" is complete and takes one line.

How Many Databases is Enough

9 September 20264 min

Two failure modes appear repeatedly in submitted reviews. One team searches PubMed and Google Scholar, and the search is rejected as inadequate. Another searches eleven databases including three with no relevant coverage, and reports 40,000 records screened, most of them noise.

Neither is a defensible design, and the difference between them is not the count.

What the evidence supports

Bramer and colleagues (2017) examined which database combinations achieve adequate recall across a large set of published reviews. Their finding was that a core combination of Embase, MEDLINE, Web of Science Core Collection and Google Scholar retrieved an acceptable proportion of included studies for most topics, with CENTRAL added for reviews of interventions.

The important part is that adequacy came from the combination rather than the number. Embase and MEDLINE overlap substantially but not completely; Embase indexes more European and pharmacological literature, and reviews restricted to MEDLINE miss studies that would have changed pooled estimates.

Halladay and colleagues (2015) approached it from the other end, finding that sources beyond PubMed contributed relatively few additional included studies in the reviews they examined, though the contribution was not zero. Read together, the two results say something reasonable: the marginal return on each additional database falls steeply, and the first few are not optional.

A defensible minimum by review type

For a review of health interventions: MEDLINE, Embase, CENTRAL, plus one subject-specific database. For exercise, rehabilitation and sports science that usually means SPORTDiscus or CINAHL; for psychology, PsycINFO; for nursing, CINAHL.

For a review including non-randomised or observational evidence: as above, with Web of Science or Scopus for citation coverage.

For a scoping review with a broad question: the same core, and PRISMA-ScR expects the search to be reported in sufficient detail to be reproduced, which matters more than the count.

Add grey literature where publication bias is a plausible concern: trial registries, dissertation databases, and relevant organisational websites. Trial registry searching also supports the assessment of selective outcome reporting, which is a second reason to do it.

What reviewers actually check

Not the number of databases. Four things.

Whether the full search strategy for at least one database appears in full, with every line, including limits and the exact date run. PRISMA-S is the reporting standard here and the majority of published reviews still do not meet it.

Whether the strategy was translated properly across platforms. MeSH terms do not exist in Embase, which uses Emtree, and a strategy copied unchanged from one to the other silently loses records. Automation tools for cross-database translation are among the few search-support approaches with any supporting evidence.

Whether a subject-specific database was included where one obviously applies. A review of resistance training that does not search SPORTDiscus invites the question.

Whether the search was updated. If more than 12 months separate the search date and submission, expect to be asked to re-run it, and plan for that rather than being surprised.

Where the effort is better spent

Search quality is mostly a function of the strategy rather than the platform count. Three practices with more return than a fifth database.

Build the strategy from seed studies. Take five to ten papers you know should be retrieved, examine their indexing terms and title words, and construct the strategy from that vocabulary. Then test whether the finished strategy retrieves all of them. This is one of the approaches a 2026 systematic review of search methods identified as likely to improve searches, alongside study-design filters, backwards citation searching, and automation for translating strategies across databases.

Do backwards citation searching on the included studies. It is quick and it catches records that indexing missed.

Have the strategy peer reviewed, ideally against PRESS. An information specialist will find errors in an afternoon that would otherwise become a reviewer comment.

References

Bramer, W. M., Rethlefsen, M. L., Kleijnen, J., & Franco, O. H. (2017). Optimal database combinations for literature searches in systematic reviews: A prospective exploratory study. Systematic Reviews, 6, 245. https://doi.org/10.1186/s13643-017-0644-y

Halladay, C. W., Trikalinos, T. A., Schmid, I. T., Schmid, C. H., & Dahabreh, I. J. (2015). Using data sources beyond PubMed has a modest impact on the results of systematic reviews of therapeutic interventions. Journal of Clinical Epidemiology, 68(9), 1076–1084. https://doi.org/10.1016/j.jclinepi.2014.12.017

Common questions

Is Google Scholar an acceptable database for a systematic review?: Not as a primary source. Its results are not reproducible, the interface caps retrieval, and the search syntax is limited. It is useful as a supplementary check and for grey literature, and it should be reported as such, with the date and the number of results screened.
Do I have to search in languages other than English?: You do not have to, but you must report the restriction and consider its effect. In fields with substantial non-English literature, notably Chinese-language exercise and rehabilitation trials, an English-only restriction can bias the evidence base. If you restrict, say so in the limitations and explain why.
How do I report a search that was updated during the review?: Report both dates and what the update returned. A sentence such as "The search was run on 3 February 2026 and updated on 18 August 2026; the update identified 412 additional records, of which 3 met the inclusion criteria and were added" is complete and takes one line.