What a thesis literature review actually involves
It is not a longer essay. Four things separate it from undergraduate work, and they are all about time.
It runs for months. You will read something in October and need to cite it precisely in June, by which point you will not remember which of four similar papers said the thing you are attributing.
It has to be defensible. In a viva, "I read that somewhere" is a bad answer. Every claim needs to be traceable to a source, and ideally to a page in that source.
The scope moves. Your question in month two is not your question in month ten, and the review has to survive that without being rebuilt from scratch.
It is assessed as a piece of reasoning. A chapter that lists studies in sequence fails even when every study is relevant. What is being marked is the synthesis — the grouping, the disagreements, the gap your work fills.
The specific problems
Attribution decays. Notes written six months ago lose their link to the source. This is the single most common way a thesis chapter becomes unverifiable, and it is a filing problem long before it is a writing problem.
Search strategies get lost. If you cannot reproduce in month nine the search you ran in month two, you cannot say what you did — and in some disciplines you will be asked to put the strategy in an appendix.
A general index is not a subject search. Wonders searches OpenAlex and Semantic Scholar; OpenAlex's own works endpoint reported 327,790,539 indexed works when we read it on 22 September 2026. Breadth is not the same as a controlled-vocabulary search: the National Library of Medicine's MeSH, "a controlled and hierarchically-organized vocabulary", governs what a PubMed query retrieves in a way no free-text search reproduces. The same is true of your field's own database. Use both.
The AI question is now a compliance question. Departments have policies, they differ, and they change. The policy is the authority, not a vendor's marketing page — including this one.
What Wonders does across those months
- Keeps one project in three tabs. Search, Organize, Write. The searches you ran, the sources you kept and the draft live in the same project rather than three applications.
- Shows the search strategy as a string. The Boolean builder displays the query it runs, so a strategy is something you can copy into an appendix and re-run later.
- Runs sub-questions side by side. One panel per topic in the Search tab, each with its own query and filters — which is the shape of a review with several strands.
- Checks quotes before showing them. If Wonders cannot find a quoted phrase in the paper's text, it does not present it as a quote. Where full text is available, a citation carries the page number, and clicking it opens the PDF at that passage. That is what makes a claim defensible in month nine.
- Weights journal quality. SCImago quartiles for more than 28,000 journals in the ranking and as a filter.
- Resolves full text through your library. OpenURL, EZproxy, OpenAthens, KBART and LibKey, where your institution has set them up.
- Writes with citations attached. Inline citations in the notebook and an automatic bibliography in APA, MLA, Chicago, Harvard, IEEE or Vancouver; export to DOCX, PDF, LaTeX or HTML, and BibTeX, RIS or CSV; save to Zotero; import
.ris,.biband.nbibfrom whatever you were using before. - Chat with one paper, after login, when a method section will not give up its point.
- Works in 16 languages in the app interface, which matters if you are writing in English but thinking in something else.
Citation provenance and the supervisor audit view, real-time co-editing and project-wide chat are in beta and not generally available.
What Wonders does not do, and what to use instead
| The job | Does Wonders do it? | What to use |
|---|---|---|
| Write the chapter or paraphrase a source | No, deliberately | You. That is the assessment |
| Detect research gaps | No | Our three-step method, applied by hand |
| Systematic review screening, dual review, PRISMA | No | A dedicated screening tool; see systematic, scoping and narrative reviews first |
| Data-extraction tables across studies | No | A spreadsheet, or a synthesis matrix |
| Search CINAHL, PsycInfo, ERIC, Scopus, Web of Science | No | Those databases, through your library |
| Pull your existing Zotero library in | No | Export it as .ris or .bib and import that. Saving out to Zotero works |
| A browser extension, or an API | No | Neither exists |
Use something else if you want the chapter drafted for you. Wonders will not do it, and a tool that does is a bad idea in a thesis you have to defend in person.
Wonders is a reasonable fit if the hard part is volume and traceability: hundreds of papers, a moving question, and an examiner who will ask where a claim came from.
Before you start
- How to start a literature review — the first two weeks.
- Plan your PhD literature review — planning across a year.
- How many sources does a literature review need? — the question everyone asks their supervisor.
- How to read a research paper: the three-pass method — how to get through the pile.
- AI-assisted research ethics — read this alongside your own university's policy.
What it costs
Plus is $20 a month, or $144 a year — $12 a month. Verified students pay $10 a month, or $72 a year — $6 a month, for up to four years. There is a 14-day free trial and no free plan. See pricing.