What Google Scholar is, and what it is not
Google Scholar is a free search engine for scholarly material. It indexes journal articles, books and chapters, theses, conference papers, preprints and working papers found on publisher sites, repositories and university pages. That breadth is its strength: it often surfaces material that subject databases miss, including working papers and books that matter in fields such as economics and business (Harzing & van der Wal, 2009), and grey literature such as reports and theses (Haddaway et al., 2015).
It is not a bibliographic database in the usual sense. Its coverage is not published, its ranking is algorithmic and undisclosed, its query syntax is limited, results cannot be exported in bulk, and the same query can return different results on different days. Those features matter little when you are looking for a known paper, but they matter a great deal when a search must be transparent and repeatable, as in a systematic review (Page et al., 2021).

Plan your keywords before you search
Too many results is the first problem most people meet. The answer is a search plan, not more scrolling. Break the research question into its main concepts, list synonyms and related terms for each, and decide which concepts must appear together. Frameworks such as PICO for clinical questions or SPIDER for qualitative evidence can help structure the concepts (Cooke et al., 2012). If a topic is unfamiliar, read an encyclopaedia entry or a recent review first to learn its vocabulary.
In Scholar, use quotation marks for exact phrases, OR between synonyms and a minus sign to exclude a term, and use Advanced Search to restrict words to the title, to a date range, to an author or to a publication. Keep a search log: the date, the exact string, the filters and the number of results. That log lets you repeat the search later and report it honestly, and it is the habit that distinguishes a systematic search from browsing (Booth et al., 2016).
Backward and forward citation searching
Citation searching is where Scholar is most useful. Backward searching means working through the reference list of a relevant paper to find earlier work. Forward searching means finding later work that cites it, which is what Scholar's "Cited by" link shows. "Cited by" is sometimes described as a backward search; it is the forward direction.
Used systematically, the two directions form "snowballing": start from a set of clearly relevant papers, screen their references and citations, add the relevant ones to the set and repeat until no new relevant papers appear (Wohlin, 2014). Snowballing complements keyword searching because it finds papers that use different terms for the same idea. "Related articles" works in a similar spirit but on an undisclosed similarity measure, so treat it as a discovery aid rather than a method you can report.
Record every search as you go
Because Scholar results change over time and its ranking is not disclosed, the only way to make a Scholar search accountable is to record it at the moment you run it. A simple log kept in a spreadsheet is enough, and it becomes the raw material for the search description in your methods section or literature review chapter. It also protects you months later, when a supervisor or examiner asks how a particular study was found, or why another was not.
| Field | Example entry |
|---|---|
| Date and source | 12 March, Google Scholar (signed in, library links on) |
| Exact string | "peer feedback" (online OR e-learning OR MOOC) |
| Filters | 2015 onwards; articles only |
| Results screened | First 10 pages by title; 23 abstracts read |
| Kept | 6 studies, saved to My Library with the label "PF-online" |
| Notes | Many hits use "peer assessment"; add as synonym in next run |
Note in the log where you stopped screening and why. With Scholar, stopping after a fixed number of pages is normal and acceptable if you say so; what examiners object to is a search that cannot be described at all. Repeat key searches near the end of the doctorate, since relevant work will have appeared in the meantime, and record the second run in the same way.
Features that save time
- Library links. Link Scholar to your university library in the settings so that full-text links appear for material your library subscribes to.
- My Library. Save papers with labels as you search, then export them to a reference manager.
- Alerts. Set an alert on a key search or on citations to a key paper so new work arrives by email during your doctorate.
- Versions. "All versions" can lead to a free repository copy of a paywalled article. Cite the version of record where one exists.
- Cite. Useful as a starting point, but check authors, year, volume, pages and DOI against the publisher's page before the reference enters your thesis.
The same caution applies more strongly to AI search and chat tools. A study of ChatGPT-generated bibliographies found that many citations were fabricated and many real ones contained errors (Walters & Wilder, 2023). Treat any reference you have not opened yourself as unverified. For managing and formatting references, see the companion guide on referencing in a PhD thesis.
Where Google Scholar falls short
| Task | Google Scholar | Bibliographic databases (e.g. Scopus, Web of Science, subject indexes) |
|---|---|---|
| Find a known paper quickly | Very good | Good |
| Forward citation searching | Good, broad coverage | Good, curated coverage |
| Grey literature, theses, working papers | Good | Limited, varies by database |
| Complex Boolean search with nesting | Limited | Good |
| Subject headings or thesaurus terms | None | Available in many subject databases |
| Bulk export of results for screening | Not available | Available |
| Reproducible, reportable search | Weak: results and ranking change | Strong |
These limits matter most in systematic and scoping reviews, where PRISMA 2020 asks authors to report every source and full search strategy (Page et al., 2021). In a prospective study of systematic review searches, Bramer et al. (2017) found that no single database retrieved all included references and that a combination of Embase, MEDLINE, Web of Science and Google Scholar performed best in their sample. Google Scholar contributed, but as one source among several. Haddaway et al. (2015) similarly concluded that Scholar is valuable for grey literature but should supplement, not replace, database searching.
Databases have their own biases. Web of Science and Scopus cover the natural sciences, engineering and biomedicine more fully than the social sciences and humanities (Mongeon & Paul-Hus, 2016), and free indexes such as OpenAlex are now being evaluated against them (Culbert et al., 2025). Grey literature matters too: in health research, meta-analyses that exclude it can overestimate intervention effects (Hopewell et al., 2007). The practical lesson is to choose sources for your field and question, not by habit.
How much searching does your review need?
The searching you need depends on the kind of review you are writing (Grant & Booth, 2009). A narrative literature review chapter needs a search that is thorough and explained; a systematic review needs one that is comprehensive and reproducible.
- Background reading and the first draft of a chapter: Scholar plus citation searching is often enough to map the field.
- A thesis literature review chapter: add at least one major subject database for your discipline, keep a search log and describe your approach in a short paragraph.
- A systematic or scoping review: search several databases with full strategies, add Scholar and other grey literature sources as supplements, and report everything. See searching the psychology literature for a database-led example.
Worked example (illustrative)
A doctoral researcher in education is reviewing research on peer feedback in online courses. She begins in Scholar with "peer feedback" AND ("online learning" OR "e-learning" OR MOOC), restricted to 2015 onwards, and logs the string and date. From the first pages she selects five clearly relevant studies. Backward searching of their reference lists adds eleven papers; forward searching through "Cited by" adds nine more recent ones, including two doctoral theses. She then runs a structured search in an education database using its thesaurus terms, which finds studies that call the same practice "peer assessment". Finally she sets Scholar alerts on two key papers. Her methods paragraph reports the databases, strings, dates and the snowballing step, so an examiner can see how the evidence base was built.
For turning the material into an argument, see how to write a literature review for a PhD and how to identify a research gap.
Common mistakes
- Relying on Scholar alone for a systematic review.
- Calling "Cited by" a backward search, or using only one direction of citation searching.
- Not recording search strings, dates and filters, so the search cannot be repeated or reported.
- Copying the "Cite" output or an AI-suggested reference without checking the source.
- Treating "Related articles" or result counts as evidence of completeness.
- Searching only in English and only in one discipline's vocabulary.
Key takeaways
- Google Scholar is broad and free and is strongest for known-item searching, citation chasing and grey literature (Haddaway et al., 2015).
- "Cited by" is forward citation searching; combine it with backward searching in a deliberate snowballing process (Wohlin, 2014).
- For systematic reviews, no single source is sufficient; use several databases and report every search (Bramer et al., 2017; Page et al., 2021).
- Choose databases that cover your discipline, since coverage differs between indexes (Mongeon & Paul-Hus, 2016).
- Verify every reference from Scholar or an AI tool against the original before citing it (Walters & Wilder, 2023).
