A draft literature review lands in your lap — a chapter from a departing student, a section a supervisor handed back with "please finish this," or a file left in a shared project folder over winter break. The prose reads smoothly. The citations look plausible. Then you try to extend it, and you realize there is no record of how any of it was found: no query strings, no database names, no dates, no screening notes, no list of what was rejected. You have inherited conclusions without their provenance.
This is a common and fixable situation. The repair is not to rewrite the prose first. It is to reconstruct the search behind it, decide what the review is actually claiming, and rebuild the missing artifacts so the next person — likely you — can extend and defend it. This guide walks through diagnosing your starting state, making the key decisions, and knowing when the rebuilt review is ready to hand on.
What does "no search history" actually cost you?
A literature review without a search record is hard to trust and hard to extend, because the reader cannot tell whether the included studies were found systematically or assembled by convenience. Search methodology guidance treats the search as a designed, documented process: Cochrane's handbook, for instance, recommends planning searches from explicit eligibility criteria, drawing on multiple sources, combining free-text terms with subject headings, documenting the process, deduplicating records, and screening in stages. When that record is missing, each of those steps has to be reconstructed or re-run.
The practical costs show up in four places:
- Coverage. You cannot tell which databases, sources, or date ranges were consulted, so you cannot say what the review might have missed.
- Reproducibility. A reader or examiner cannot re-run the search to check whether it still returns the same core set.
- Screening logic. Without inclusion and exclusion decisions on record, borderline papers are impossible to adjudicate consistently.
- Extension. Adding a new year of literature means re-deriving the original strategy from scratch.
None of these require you to discard the draft. They require you to treat the existing prose as a hypothesis about the field and rebuild the evidence trail underneath it.
Diagnose your starting state before you touch the prose
Before rebuilding anything, work out what you actually have. Inherited reviews fall into a few recognizable states, and each needs a different repair.
The four common inherited states
- Prose only. A written narrative with citations but no notes, no database list, and no screening record. This is the most common inheritance and the most work to repair.
- A reference list with no method. Citations exist in a bibliography, but nothing explains how they were selected or whether the list is complete.
- Partial notes. Some search terms or a few saved PDFs survive, but they are inconsistent and undated.
- A method section that describes a search nobody can reproduce. The draft claims a systematic approach but gives no terms, sources, or dates that would let you repeat it.
A short diagnostic checklist
Ask these questions of the inherited file and write down the answers:
- What is the review's stated question or scope, if any?
- Which sources appear to have been used, judging from the citation patterns?
- Is there any record of search terms, dates, or filters?
- Are inclusion and exclusion criteria written down anywhere?
- Which citations are load-bearing — the ones the argument depends on — versus decorative?
- What time period does the review claim to cover, and where does it stop?
If most answers are "unknown," treat the draft as prose-only and plan a full rebuild of the search layer. If some notes survive, you can validate and extend them rather than starting cold.
Decide what the review is for before rebuilding the search
The single most useful decision is what the review needs to do. A course paper, a thesis chapter, and a first manuscript have different burdens.
- Coursework usually needs defensible coverage and clear reasoning, not an exhaustive systematic search.
- A thesis chapter needs a search you can describe and defend in a methods section, because an examiner may ask how you found your sources.
- A first manuscript needs a search record that supports the claims in the introduction and related-work sections, and that you can update before submission.
Write one sentence stating the review's purpose and scope. That sentence becomes the anchor for every later decision about what to include. Cochrane's guidance to plan searches from eligibility criteria is the principle here: define what counts as relevant before you go looking, so the rebuilt search has a target.
Reconstruct the search strategy from the citations you already have
You are not starting from zero. The inherited citations are clues. Use them to reverse-engineer the search that probably produced them.
Mine the existing citations for terms and sources
Take the load-bearing citations and extract:
- the key concepts and synonyms in their titles and abstracts;
- the journals and venues they come from;
- the authors who recur across the list;
- the date range the citations cluster in.
These give you a starter set of free-text terms and a sense of which sources matter. Google Scholar's documented features are useful here: you can search by author and title, sort and limit by date, follow "Cited by" and "Related articles" links, and locate alternative versions of a work. Following citations forward and backward from a known good paper is one of the fastest ways to rebuild a candidate pool.
Turn concepts into a documented query set
Build a small, explicit query table. For each core concept, list synonyms and, where a database supports them, subject headings. PubMed's documentation is a good model for what to record: field tags, phrase searching, filters, and the search details that reveal how terms were translated and combined. Record the exact strings you run, the source, the date, and any filters applied.
When you need to find papers through a scholarly search interface, WisPaper search gives you a place to run those queries and collect results; the public Scholar Search page explains the entry point for finding academic papers. Keep your query table alongside whatever tool you use, because the table — not the tool — is the artifact that survives.
Rebuild the screening record in stages
A search without a screening record is only half a method. Cochrane recommends deduplicating records and screening in stages, and linking multiple reports of the same study. Reconstruct that logic explicitly.
Set inclusion and exclusion criteria now
Write down, in plain language, what makes a paper eligible: population, intervention or topic, study type, date range, language, and any other boundary. These criteria let you and a collaborator make the same decision about a borderline paper.
Screen in two passes
First pass on titles and abstracts, second pass on full text. For each excluded paper at full-text stage, note the reason. This record is what turns a pile of citations into a defensible method.
Handle duplicates and linked reports
Deduplicate before screening so the same study is not counted twice. Where several reports describe one study, link them and count the study once. Both steps are standard in systematic search guidance and both are easy to skip when you are rebuilding quickly.
As you assemble the candidate pool, save the papers you intend to keep. My Library lets you save or upload papers and organize them in one place, which gives the rebuilt review a stable home for its included set.
Verify the citations you inherited
Inherited reference lists deserve a check, especially when the draft passed through several hands. A citation can be misattributed, duplicated, or simply wrong. Before you build on a reference, confirm it exists and says what the draft claims.
For BibTeX-based reference lists, TrueCite checks BibTeX entries against real academic databases to flag hallucinated references. Run your inherited bibliography through a check like this early, so you are not repairing the search around citations that should not be there at all.
Re-run and update the search before you call it finished
Cochrane recommends rerunning searches before publication. For an inherited review, this step does double duty: it validates your reconstructed strategy and it brings the review up to date.
Re-run your documented queries, note the date, and compare the new results against your included set. If the strategy is sound, the re-run should surface the core papers you already have plus any recent additions. If it surfaces a large set of relevant papers you missed, your query set needs revision — which is exactly the information the original review lacked.
Organize the rebuild so the next person can extend it
A repaired review should leave behind artifacts, not just better prose. Set up a workspace that holds the query table, the screening log, the included set, and the draft together.
Research Projects gives you a place to create a project and keep the work organized as it grows. For longer reconstruction tasks — rebuilding a full review, mapping a field, or planning how to extend the draft — Scholar Agent supports literature review and related research workflows. When you need to ask questions of the papers you have already collected, Library QA lets you query papers uploaded to My Library or added from search results, which is useful for checking what the included set actually says before you rewrite the synthesis.
If the repaired review is meant to seed your next project rather than stand alone, Idea Discovery takes research topics, background, references, and experimental constraints and generates research ideas — a natural next step once the literature base is trustworthy.
Know when the rebuilt review is ready
The review is ready when the artifacts, not just the prose, are complete. Use this checklist:
- A one-sentence statement of the review's purpose and scope.
- A documented query table with terms, sources, dates, and filters.
- Written inclusion and exclusion criteria.
- A screening log with reasons for full-text exclusions.
- A deduplicated included set with linked reports of the same study.
- A verified reference list.
- A dated re-run of the search showing the strategy still works.
- A synthesis that matches the included set — no citations in the prose that are absent from the record.
When all eight are true, you can hand the review to a supervisor, an examiner, or a co-author and they can see how it was built. That is the difference between inherited prose and a review you own.
A winter-break cleanup, start to finish
Suppose a master's student inherits a related-work chapter from a student who graduated in December. The chapter has forty citations and no method section. Over winter break, the student:
- Reads the chapter and marks the ten load-bearing citations.
- Extracts concepts, synonyms, venues, and the date range from those ten.
- Builds a query table and runs it across the sources the citations suggest, saving results in My Library.
- Writes inclusion and exclusion criteria and screens in two passes, logging exclusions.
- Checks the inherited BibTeX with TrueCite and removes references that do not verify.
- Re-runs the queries in January, notes the date, and adds any new relevant papers.
- Rewrites the synthesis against the included set and adds a short method paragraph describing the rebuilt search.
The chapter now has a search history. The student also has a query table and screening log they can reuse for their own thesis.
Transferable lessons for your next project
- Document as you search, not after. The query table and screening log are cheap to maintain and expensive to reconstruct.
- Define eligibility before searching. Criteria first, then queries.
- Keep the artifacts with the draft. A review is a method plus a synthesis, not prose alone.
- Re-run before you submit. A dated re-run is the simplest evidence that your strategy still works.
- Verify inherited references early. It is easier to rebuild around a clean bibliography than a contaminated one.




