The best AI tools for PhD students each cover one stage of research: Semantic Scholar and Consensus to find relevant papers and evidence, ResearchRabbit to map citation networks, Elicit to extract data from many papers, Zotero to manage references, NotebookLM to ask questions across your own PDFs, and Type to Study to actually learn and remember a new field. No single tool does all of it, and none replaces reading the key papers yourself.
What stage of research are you at?
Researchers often pick tools by popularity instead of by task. A better approach is to name the job first. Finding papers, understanding a field, extracting results, managing citations and writing are different problems, and the tools below are good at different ones.
| Tool | Main job | Good to know |
|---|---|---|
| Semantic Scholar | Search for papers, with short AI summaries and citation information | Free, run by the Allen Institute for AI |
| Consensus | Answer a specific research question from published studies | Best for quick evidence checks, not a full review |
| ResearchRabbit | Discover related papers through citation maps | Visual; good for finding what you missed |
| Elicit | Find papers and extract structured data into tables | Useful for systematic and scoping reviews; check extractions |
| Zotero | Store, organise and cite references | Free and open source; works with word processors |
| NotebookLM | Ask questions across PDFs you upload, with citations | Answers only from your sources |
| Type to Study | Learn a field and retain it: primer, lessons, recall, flashcards, test | A learning tool, not a reference manager or search engine |
Finding papers: Semantic Scholar and Consensus
Semantic Scholar is a free academic search engine that adds short summaries of papers and shows how they are cited. Consensus takes a question, such as whether an intervention improves an outcome, and summarises what published studies say. Both speed up the first pass. Neither tells you which studies are well designed; that judgement is still yours.
Mapping a literature: ResearchRabbit
Once you have a few key papers, ResearchRabbit builds maps of papers that cite them or are cited by them, and suggests related work. It is good for spotting clusters and authors you had not found. Litmaps does similar citation mapping if you prefer its layout.
Extracting results: Elicit
Elicit is built for reviews. It finds papers and pulls details such as sample size, methods and outcomes into a table you can scan. That saves time on screening, but extraction can be wrong, so check values against the paper before you report them.
Managing references: Zotero
Zotero is not an AI tool, but every PhD student should use a reference manager from day one. It saves papers from your browser, keeps PDFs and notes together, and inserts citations in your thesis. Starting late means hours of tidying later.
Questioning your own sources: NotebookLM
Upload a set of papers and ask NotebookLM what they say about a method or a disagreement, and it answers with citations to the passages. It is good for orientation within sources you have chosen. Reading its answer is still passive, so you will not remember much of it next month.
Learning a new field and keeping it: Type to Study
Starting a PhD, switching fields or joining a new lab means learning a body of knowledge fast: terms, methods, landmark findings and debates. Type to Study is built for that. Type a research topic and it builds a primer from research papers found through OpenAlex, encyclopedias, web articles and books, with every source listed. Or upload the key papers and review articles as PDFs and it turns them into units and lessons that cite the pages.
You read each lesson, type the key definitions and findings, then do a recall pass where the key terms are hidden. Terms become spaced flashcards, and a final test shows which parts of the field you can explain and which you cannot. The AI tutor answers from your own lessons with citations. It does not search the literature exhaustively, manage references or write for you.
A workflow for your first month in a new field
- Search Semantic Scholar or Consensus for two or three recent reviews.
- Map from those reviews in ResearchRabbit to find the landmark papers.
- Save everything in Zotero with tags.
- Upload the reviews to Type to Study and work through the lessons and recall passes, 20 to 30 minutes a day.
- Use NotebookLM or the tutor to ask about points of disagreement, then read those sections of the papers yourself.
- Take the final test before your first supervisor meeting and revise what it flags.
Questions researchers ask
Is using AI tools allowed in a PhD?
Policies differ by university and journal. Using tools to find and understand papers is widely accepted; using them to write text you submit is often restricted. Check your institution's guidance and disclose use where required.
Can AI do my literature review?
It can speed up searching, screening and extraction. It cannot judge study quality or build your argument, and it can miss or misread papers. The review remains your work.
Which tool should I start with?
Zotero first, because it pays off for years. Then a search tool, then whichever of the others matches your current bottleneck.