What Should You Record So You Can Reproduce Your Literature Search Later?

What Will You Need Months Later to Recreate the Search You Ran Today?

You finish a literature search and save the articles you found. Several months later, you need to explain exactly how you searched.

You remember using Scopus and perhaps Web of Science. You remember searching terms related to generative AI, academic writing, and university students. You probably used quotation marks somewhere. You may have changed the query several times before settling on the final version.

But which version was actually run? On what date? Which database platform did you use? What fields did you search? Were any year, language, document-type, or subject limits active? How many records did the final search retrieve? Did you search another source later? What changed when you updated the search?

If those details were not recorded at the time, reconstructing them from memory can be surprisingly difficult.

A reproducible literature search therefore requires more than preserving the papers you found. You need a contemporaneous record of what you searched, where you searched it, when you searched it, how the query was constructed, what restrictions were applied, and what the search produced.

Record Enough Information for Someone Else to Rerun the Search

A Reproducible Search Is a Record of Actions, Not a Summary From Memory

Documentation and reporting are related but not identical

It helps to distinguish what you record while conducting the search from what eventually appears in the published paper, thesis, protocol, or review.

Cochrane makes this distinction explicitly. Its current Handbook describes documentation as internal record-keeping and requires the search process to be documented in enough detail that it can later be reported correctly and the searches can be reproduced to the extent possible.

Good reporting therefore begins long before you write the methods section.

Record the exact database, not merely the general resource name

“We searched PubMed, Scopus, and other databases” is not sufficient documentation for a systematic search.

Record the exact source searched. Where relevant, distinguish the database from the interface or platform through which you accessed it.

For example, MEDLINE can be searched through different platforms. The underlying bibliographic database and the interface are not necessarily interchangeable because search fields, syntax, controlled vocabulary functions, proximity operators, and other features can differ.

PRISMA 2020 requires systematic reviews to specify all databases, registers, websites, organizations, reference lists, and other sources searched or consulted, as well as the date when each was last searched. Its explanation and elaboration also recommends identifying database interfaces or platforms where relevant.

Record the date each source was actually searched

A database search is not timeless.

Databases continually add records, update indexing, correct metadata, and change coverage. Running the same strategy several months later can therefore produce a different result count even when the query itself has not changed.

Record the actual search date for every source.

If you later update the search, record the new date separately rather than replacing the original date. This creates a history of what evidence could have been retrieved at each stage.

These are different pieces of information and may both be necessary.

Save the complete search strategy exactly as it was run

This is probably the most important habit to develop.

Do not save only a simplified description such as:

Keywords: artificial intelligence, academic writing, students.

That does not reveal how the terms were combined, which phrases were searched, which fields were used, whether controlled vocabulary was included, or what syntax the database interpreted.

Instead, preserve the complete strategy.

For a structured bibliographic database search, that may include:

• every search line;
• all free-text terms;
• all controlled vocabulary terms;
• Boolean operators;
• parentheses and nesting;
• phrase syntax;
• truncation and wildcard symbols;
• proximity operators;
• field codes;
• search set numbers;
• filters and limits.

PRISMA 2020 recommends presenting the full search strategies for all databases, registers, and websites searched, including filters and limits. Cochrane similarly requires database strategies to be preserved exactly as run.

Copy and paste the strategy rather than retyping it

Cochrane specifically advises copying and pasting bibliographic database search strategies exactly as run rather than retyping them because transcription can introduce errors.

This is a deceptively important recommendation.

A search containing dozens of lines, nested Boolean expressions, subject headings, proximity operators, and field codes is easy to alter accidentally. A missing parenthesis or mistyped truncation symbol may produce a strategy different from the one that generated the reported results.

Save the original output from the search platform whenever possible.

Many interfaces allow search histories to be saved, exported, emailed, or downloaded. An offline text copy is also useful because institutional accounts, subscriptions, interfaces, and saved-search features can change.

Record every limit and filter that affected retrieval

A complete search string is still incomplete documentation if additional filters were applied through the database interface.

Record restrictions involving:

• publication date;
• language;
• document or publication type;
• peer-review status;
• age or population filters;
• subject categories;
• study-design filters;
• other database-specific limits.

PRISMA 2020 explicitly requires full strategies to include filters and limits. Cochrane likewise expects restrictions to be documented and reported.

This matters because two researchers can enter an identical text query and obtain different results if one has active filters that the other does not know about.

Several of these restrictions also involve substantive methodological choices. Decisions about date ranges, language restrictions, and whether to search only peer-reviewed literature should therefore be documented alongside their implementation.

Record the number of results retrieved from each search

The result count provides an important checkpoint.

Suppose you rerun a saved strategy six months later and retrieve 1,842 records rather than the original 1,615. That difference may be entirely expected because the database has added records. But you need the original count to know that anything changed.

Cochrane requires the total number of records retrieved by each bibliographic database strategy to be preserved with the strategy. Result counts also support later construction of the study-selection flow diagram.

For each source, record at least the count associated with the final strategy actually used for evidence identification.

If the source is a web search engine where thousands of results are returned but only a limited number are screened, document the number actually screened when that better represents what was examined.

Do not discard the development history too quickly

During search development, you may test several versions:

You do not necessarily need to publish every experimental query you tried. But preserving consequential revisions and their rationale can be extremely useful, particularly for complex systematic searches.

If a reviewer later asks why an outcome was omitted or why a term was added, you have a contemporaneous record rather than an improvised explanation.

This is particularly valuable when you have iteratively adjusted the breadth of the initial search or corrected a strategy that was too narrow.

Record who performed the search

Cochrane's documentation standard explicitly includes who conducted the search alongside the sources, dates, and terms used.

This information can matter when several researchers, librarians, or information specialists contribute to different parts of the process.

For example, one team member may develop the MEDLINE strategy, an information specialist may translate it to Embase, and another researcher may conduct citation searching. Knowing who performed each activity helps resolve questions later and supports transparent methods reporting.

Record searches outside bibliographic databases too

A reproducible search log should not stop with Scopus, MEDLINE, ERIC, or other conventional databases.

If you search trial registers, organizational websites, repositories, conference proceedings, regulatory sources, Google Scholar, or other resources, document those activities as well.

Depending on the source, record:

• the exact resource or website;
• the date searched;
• the search terms or query used;
• the interface or search mode where relevant;
• filters or limits;
• the number of records or results screened;
• what portion of the results was examined if the source could not practically be screened in full.

Cochrane emphasizes that searching sources other than bibliographic databases also needs to be documented when reproducibility is sought. Web searches may not be reproducible to the same degree because content, algorithms, personalization, and ranking can change, but documenting what was done remains essential.

Record backward and forward citation searching

Citation searching can become surprisingly difficult to describe after the fact.

If you examine references cited by relevant publications, record which publications were used as the starting points and when the searching occurred.

If you conduct forward citation searching, also record the citation index or platform used.

PRISMA 2020 specifically recommends reporting the bibliographic details of reports used for backward or forward citation searching, the citation index or platform, and the date the search was conducted.

This is more informative than writing simply:

Reference lists were also searched.

Which reference lists? At what stage? Were all included studies checked, only review articles, or a few seed papers? Was forward citation searching performed as well?

The method should be reconstructable.

Record handsearching and browsing activities

If you manually inspect journals, conference proceedings, websites, or repository collections, preserve the scope of that activity.

For a journal, for example, record the journal name, date or volume range covered, date searched, and whether you browsed tables of contents, searched the journal website, or used another method.

PRISMA 2020 recommends specifying the names of journals or conference proceedings consulted, the dates covered, and how they were searched.

“Handsearched relevant journals” is too vague to reproduce.

Record contact with experts, authors, and organizations

If you contact researchers, study authors, manufacturers, organizations, or other experts to identify additional studies or obtain information, record that activity.

Useful details can include:

• who or what organization was contacted;
• the date of contact;
• the purpose of the request;
• what information or studies were requested;
• whether a response was received;
• whether additional records or study information resulted.

Not all of this detail necessarily belongs in the main article, but retaining it allows the final methods and supplementary materials to be accurate.

Keep database records separate long enough to preserve source provenance

When results from several databases are immediately merged into one reference library, it can become difficult to determine where each record originated.

That information matters for documenting the flow of records and evaluating source contributions.

A practical workflow is to preserve the source of each imported batch, whether through separate files, database labels, import groups, custom fields, or systematic filenames, before deduplication changes the record set.

For example:

2026-08-30_Scopus_1842_records.ris

is considerably more informative six months later than:

export (7).ris

The exact file-naming system is up to you. The principle is to preserve provenance.

Record how duplicates were handled

When multiple databases are searched, duplicate records are inevitable.

For systematic reviews, the number of records before and after deduplication affects the study-selection flow. PRISMA 2020 requires reporting the records identified and those removed before screening, including duplicates.

Your internal record should therefore preserve enough information to determine:

• how many records were imported from each source;
• how many records existed before deduplication;
• what software or process was used to identify duplicates;
• how many duplicates were removed;
• how many records remained for screening.

Different deduplication tools and algorithms can produce different results, so the method can matter when rigorous reproducibility is required.

Keep records and studies conceptually separate

One study can have multiple reports.

A conference abstract, trial registration, dissertation, preprint, and journal article may all describe the same underlying investigation. If they are treated as separate studies, the evidence can be counted more than once.

Cochrane emphasizes that the unit of interest is the study rather than the report and that review authors need to link multiple records or reports to their underlying studies.

Your documentation system should therefore eventually be able to distinguish:

This becomes particularly important after you search beyond peer-reviewed journal literature, because the same study may appear through several publication routes.

Save the original exports when feasible

After importing records into a reference manager or systematic-review platform, keep the original export files where feasible.

These files provide a snapshot of what the database returned at the time of searching. They can be useful if records are accidentally altered, source provenance is lost, deduplication needs to be checked, or the flow counts need to be reconstructed.

Use systematic filenames that include the source and date. If several searches were conducted in the same source, include enough information to distinguish them.

A simple folder structure can save hours of archaeological work later.

Record search updates as new search events

If the literature search is updated before publication, do not overwrite the original strategy and pretend that only the latest search occurred.

Record:

• the original search date;
• the update date;
• the period covered by the update;
• whether the strategy changed;
• why it changed if applicable;
• the number of new records retrieved;
• how the new records were integrated and screened.

PRISMA 2020 requires the date each information source was last searched, while preserving the search history internally makes the update itself auditable.

For systematic reviews, updates can be methodologically important because databases, indexing, search interfaces, and terminology may have changed since the original search.

Do not rely only on a saved-search account

Many databases allow searches to be saved within a personal account. Use that feature when helpful, but do not make it your only record.

Accounts expire. Institutional subscriptions change. Database interfaces are redesigned. Saved histories can disappear when vendors migrate platforms.

Cochrane recommends maintaining an offline text copy of search strategies even when database interfaces allow searches to be saved or emailed.

For some resources, screenshots can provide supplementary evidence of how a search was configured. Screenshots should complement, not replace, machine-readable or text documentation when the complete strategy can be preserved directly.

A search log can keep all of this manageable

Documentation sounds burdensome when imagined as something reconstructed after the search. It is much easier when recorded as each search is run.

A basic search log might contain:

For a complex systematic review, you may need additional fields. For a conventional thesis literature search, a simpler version may be sufficient.

The principle remains the same: record information while it is cheap to capture.

Reproducible does not always mean identical results forever

Even perfect documentation cannot guarantee that another researcher will obtain exactly the same result count years later.

Databases change. New records are added. Existing records are corrected. Subject headings are updated. Journal coverage changes. Search interfaces and algorithms evolve. Websites disappear.

Reproducibility in literature searching therefore means documenting the procedure sufficiently for another researcher to understand and rerun what you did as faithfully as the current information environment permits.

Cochrane explicitly qualifies reproducibility in this way, requiring searches to be reproducible to the extent possible.

This is also why the date of searching matters so much. It anchors the search to a particular state of a changing information source.

Search documentation also makes future updates easier

A well-documented search is not useful only for peer reviewers.

It becomes the starting point when the review needs to be updated.

If you preserved the exact strategies, databases, platforms, dates, coverage, and rationale for major decisions, you can assess what still works and what needs adaptation. Without those records, updating can become a partial reconstruction of the original project.

Good documentation is therefore part of research maintenance as much as research reporting.

For systematic reviews, PRISMA-S provides a more detailed reporting framework

PRISMA 2020 establishes the core expectation that authors identify all information sources and provide full search strategies. PRISMA-S extends this with search-specific reporting guidance intended to improve transparency and reproducibility.

Its recommendations address information sources and methods, search strategies, peer review of searches, managing records, and reporting the search process.

If you are conducting a formal systematic review, consulting PRISMA-S while designing your search log is considerably easier than discovering its reporting expectations when the manuscript is nearly finished.

What Does a Reproducible Search Record Actually Look Like?

When the methods section and supplementary search appendix are eventually written, most of the necessary information already exists. The researcher is reporting a record of what happened rather than trying to reconstruct a methodological memory.

Common Mistakes When Documenting Literature Searches

Document the Search While You Still Know Exactly What You Did

The most effective search-documentation system is not necessarily elaborate. It is one that you actually update every time you search.

Capture the information at the moment it is easiest to obtain.

For a casual exploratory search, you may not need this level of detail. For a thesis, evidence synthesis, or any project in which the search method will be scrutinized, a structured search log is a modest investment with a rather impressive return when the methods chapter arrives.

What to Record Every Time You Run a Literature Search

Questions About Recording and Reproducing Literature Searches

Save the Search, Not Just What the Search Found

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.