The literature you have to search has grown nearly six times since 1995. Searching it the way your supervisor did will not work

In 1995 the world published about 2.56 million scholarly works. In 2025 it published 14.67 million — figures we took from OpenAlex yesterday. The literature a doctoral student is expected to command has grown nearly six-fold in thirty years, while the method most students use to search it has not changed at all: type some words into one search box, read what appears, stop when it feels like enough.

That method was defensible when the corpus was a fifth of its present size. It is now the most common reason a viva goes badly — not because the student has read too little, but because they cannot say what they searched, where, or what they might have missed.

What follows is the search half of the problem. The access half — how to legally open a paper once you have found it — we covered separately in reading journals from home, and it is a smaller problem than most students think.

Write the search string, then write it down

A search is a method, and a method has to be reproducible. Before running anything, draft the string.

Break the question into its two or three core concepts. For each concept, list every term an author might have used — synonyms, British and American spellings, older terminology, the abbreviation and the expansion. Join the alternatives with OR, join the concepts with AND, put multi-word terms in quotation marks, and use truncation (catalys* catching catalysis, catalyst, catalytic) where the database supports it.

Where a controlled vocabulary exists, use it. In biomedicine that means MeSH terms in PubMed, which capture papers whose authors chose different words from yours. Most indexes also allow field tags — restricting to title and abstract rather than full text — which cuts a great deal of noise.

Then record it: the exact string, the database, any date or language limits, and the date you ran it. That record is a paragraph of your methods chapter, and it is also what lets you re-run the search in eighteen months and see only what is new.

One database is not a search

The single most common flaw in an Indian literature review is that it rests entirely on Google Scholar. Scholar is a good tool with wide reach and no quality filter, no controlled vocabulary, and no way to document what it did.

Use at least three sources of different kinds:

A citation index — Scopus or Web of Science, if your institution subscribes. These are the ones that let you do forward citation searching properly. A subject index — PubMed for biomedicine, and the equivalent for your discipline; these carry the controlled vocabulary. An open index — Semantic Scholar, OpenAlex or CORE, which are free and cover a great deal that is not in the subscription indexes.

And, for anyone working on an Indian question, Indian sources explicitly. We measured this yesterday: 46 per cent of 2025 work with an Indian affiliation is closed, against 32 per cent globally, and Indian journals are indexed thinly or not at all by the international databases. Shodhganga for doctoral theses, Krishikosh for the agricultural sciences, the Indian Academy of Sciences journals, IndianJournals.com, and your discipline's Indian society journals. A review of an Indian problem built only from Scopus has a hole in exactly the place that matters.

Citation chaining finds what keywords cannot

Keyword search fails when the field uses vocabulary you have not thought of. Citation chaining does not care about vocabulary at all.

Backward: take three or four papers you know are central and read their reference lists. Anything cited by all of them is foundational, and you will not have found several of those by searching.

Forward: find everything that has cited those papers since. This is how you get from a 2011 landmark to the current state of the field in one step, and it is the function that citation indexes exist for. Google Scholar's "Cited by" does it free; Scopus and Web of Science do it better.

Tools that draw the neighbourhood — Connected Papers, Research Rabbit, Semantic Scholar — are genuinely useful here, because the underlying data is citation links rather than generated text. Use them to see the shape of a literature, not to summarise it.

The grey literature examiners ask about

Journal articles are the visible part. A doctoral committee will ask about the rest.

Theses — Shodhganga for Indian doctoral work. Read three in your own area before writing a word of yours; it teaches scope and structure faster than any guide. Conference proceedings, which in computer science and several engineering fields carry the primary literature rather than supplementary material. Standards and technical reports, which are frequently the actual specification your work must meet. Patents, where a good deal of applied work is disclosed and never published as a paper. Government reports and datasets — data.gov.in, ministry publications, statistical releases — which for policy, agriculture and public health work are primary sources.

Set the alerts in year one, not year three

This is the step that separates a literature review that stays current from one that has to be redone before submission.

Save your search in each database and turn on email alerts for new results. Set table-of-contents alerts for the five or six journals that matter most in your area. Set citation alerts on the three or four papers your work builds on, so you hear when someone extends or contradicts them. Create an ORCID and keep it current, which costs ten minutes and makes your own output findable.

The whole configuration takes an afternoon in your first term and then runs for the rest of the degree without further effort.

A reference manager, from the first paper

Use one from day one. Zotero is free, open source, and stores the PDF, the metadata and your own notes together; it produces a bibliography in whatever style is demanded, and re-styles it when the requirement changes.

Two habits make it worth far more than the setup cost. Capture the DOI with every item, because that is the identifier everything else resolves from. And write your note at the time of reading — one honest paragraph on what the paper found and why it matters to your argument. The cost of not doing this is a third year spent re-reading things you have already read because you cannot remember which one had the number you need.

The examiner's question is almost never "have you read enough?" It is "how do you know you have not missed something?" A recorded search string, several databases, citation chaining in both directions, and a note on what you excluded and why — that is the entire answer, and it is assembled while you work or not at all.

Then, and only then, the access question

Having found the papers, opening them is the smaller problem: your institution's proxy or login, the free legal version that often already exists, a repository copy, an inter-library request, or an email to the author. All of it is in our guide to reading from home, and between those routes most doctoral students can reach most of what they need.

Keeping the two problems separate matters. Students routinely narrow a literature review to what they can open, which is a search decision made by an accident of licensing. Find first, decide what is essential, then work out how to read it.

Why it matters for students and researchers

For a doctoral student, the sequence is the whole message: write the string, record it, search several places including Indian ones, chain citations both ways, set alerts, and keep everything in a reference manager with a note. None of it is difficult. All of it is much harder to do retrospectively in your final year.

For a supervisor, this is a teachable hour that saves a student a term. Most scholars are never taught to search — it is assumed to be obvious — and the gap shows up two years later as a reference list that is thin in exactly the places an examiner probes.

For an institution, the highest-return training a library can offer is a search workshop in the first month of a doctoral programme, followed by the same workshop again six months later when people have a real question to bring to it.

Frequently asked questions

How do I find research papers for a PhD in India?

Draft and record a Boolean search string, run it in at least three kinds of source — a citation index, a subject index and an open index — add Indian sources explicitly for an Indian topic, then chain citations backward and forward from the key papers.

What are the best websites to download research papers legally?

For finding: Semantic Scholar, OpenAlex, CORE, PubMed, and your institution's subscribed indexes. For opening: your library's proxy or institutional login, repository copies surfaced by Unpaywall, N-LIST if your college is a member, and inter-library requests through DELNET.

How do I do a literature review using e-journals?

Treat it as a documented method rather than a reading list: define concepts and synonyms, search multiple databases, record every string and date, deduplicate, keep the counts at each stage, and record why anything was excluded.

How do I access journals without a university subscription?

Check for a free legal version first, use NDLI and the open Indian repositories, request through a library that has inter-library arrangements, and write to corresponding authors, who may generally share their own accepted manuscripts.

Is Google Scholar enough for a PhD literature review?

No. It is a useful complement — especially its "Cited by" function — but it has no controlled vocabulary, no quality filter and no reproducible search record. Using it alone is the most common weakness in doctoral literature reviews.

What should I do in the first month of my PhD?

Create a reference manager library, an ORCID, saved-search alerts in your main databases, and table-of-contents alerts for your core journals. Then read three Shodhganga theses in your area to see what a finished thesis in your discipline actually looks like.