Build a search string for your literature review
Type your research question. The builder checks it against PICO, PEO, SPIDER or PCC, suggests synonyms, checks MeSH headings with the US National Library of Medicine, and writes the string for PubMed, Scopus, Web of Science, EBSCO, ProQuest and Google Scholar. Free, with no sign-up.
Database rules checked 4 October 2026 against each database's own help pages. For the full method, read our guide to writing a search strategy.
Search string builder
Free, no sign-up. Your question goes to our AI provider to suggest words, and your search terms go to the US National Library of Medicine and OpenAlex. We do not store them in a database or link them to you. Privacy policy
How the builder works
- It reads your question and picks a framework. An AI model from one of the providers named in our privacy policy labels the parts of your question and suggests other words for each concept. Your own words always stay in. If the model is busy or unavailable, simple rules split the question instead, and the builder tells you so.
- It adds British and American spellings, hyphenated and spaced forms, plurals of phrases and word endings (exercis* finds exercise, exercises and exercising). These follow fixed rules, and you can remove any of them.
- It looks for MeSH headings. NCBI's automatic term mapping links your words to headings, and a mapping counts only when it covers the whole term. Each heading is then looked up by name in the National Library of Medicine's MeSH vocabulary for the current year. A heading that lookup does not confirm is never shown.
- It writes one string per database from the same blocks, in that database's documented syntax: brackets around every concept, OR inside a concept, AND between concepts, and never NOT.
- It runs your blocks in OpenAlex, a free open index of scholarly works, and your PubMed string in PubMed, and shows how many records each finds. The most relevant open-access papers can go into a free Litrevu library.
- It lays out a search strategy table with the details PRISMA-S asks you to report.
Your question is not stored in a database. It goes to our AI provider to suggest synonyms (see AI processing in our privacy policy), and the words in your blocks go to the US National Library of Medicine, PubMed and OpenAlex to check headings and count results.
Check your research question: PICO, PEO, SPIDER, PCC and FINER
A framework names the parts a searchable question needs. The builder picks one from your wording and tells you which parts are missing. You can switch frameworks, and nothing you typed is lost.
| Framework | Parts | Use it when | Searched by default | Source |
|---|---|---|---|---|
| PICO | Population, Intervention, Comparison, Outcome | an intervention, treatment or programme is tested for an outcome | Population and Intervention | Cochrane Handbook, chapter 4, which credits Richardson and colleagues, 1995 |
| PEO | Population, Exposure, Outcome | an exposure, risk factor or characteristic is linked to an outcome | all three, because in an association question the outcome is part of the topic (switch it off to find more) | Moola and colleagues, 2015 (JBI) |
| SPIDER | Sample, Phenomenon of Interest, Design, Evaluation, Research type | the question is about experiences, views or perceptions (qualitative and mixed methods) | Sample and Phenomenon of Interest | Cooke, Smith and Booth, 2012; Methley and colleagues, 2014 |
| PCC | Population, Concept, Context | a scoping review maps what is known | Population and Concept | Peters and colleagues, 2020 (JBI) |
The Cochrane Handbook notes that PICO “is generally not suitable for questions around diagnostic test accuracy, prognosis, qualitative data or methods”.
FINER: is it worth asking?
- Feasible: can you read what this finds in the time you have? The results count tells you.
- Interesting: does the answer matter to you and your supervisor?
- Novel: is there a recent review on it already? The preview counts the reviews among the results.
- Ethical: would answering it need approval from your university's ethics committee? Check your own rules.
- Relevant: would the answer change practice, policy or further research in your field?
FINER is from Hulley and colleagues, Designing Clinical Research (3rd edition, 2007), as tabulated by Aslam and Emmanuel (2010).
For weak and strong examples of each framework, read how to write a research question.
What to search: the population and the intervention, not the outcome
The Cochrane Handbook says it is “usually unnecessary, however, and may even be undesirable, to search on every aspect of the review's clinical question”, because comparators and outcomes “may not be well described in the title or abstract”. Its typical strategy has three sets of terms: the population, the intervention and the types of study.
Frandsen and colleagues found that the comparison and outcome elements “had a lower retrieval potential”, and that their results “support the existing recommendation not to search for outcomes” (Journal of Clinical Epidemiology, 2020).
SPIDER needs the same care. Methley and colleagues (2014) found that “SPIDER searches showed greatest specificity for every database”, and recommended PICO “for a fully comprehensive search”.
Cochrane also says the NOT operator “should be avoided where possible”.
So the builder leaves comparison and outcome blocks off for intervention questions and leaves SPIDER's design and research-type blocks off. It never adds NOT. You can switch any block back on.
Database syntax at a glance
Checked 4 October 2026 against each database's own help pages. Every row links its source.
| Database | Title and abstract | Phrases | Truncation and wildcards | Proximity | Order without brackets | Worth knowing | Source |
|---|---|---|---|---|---|---|---|
| PubMed | term[tiab], each term tagged on its own (title, abstract and author keywords). | "phrase"[tiab] uses PubMed's phrase index. A quoted phrase that is not in the index is not searched as written; PubMed's help suggests "phrase"[tiab:~0]. | * needs at least 4 letters before it (colo*) and turns off Automatic Term Mapping. Works inside quotes ("breast feed*"). | "terms"[tiab:~N], in Title, Title/Abstract and Affiliation only. Ignored if a term has *. | Left to right. Operators in capitals. | "Heading"[mh] also finds the narrower headings under it; [mh:noexp] stops that. Shows up to 10,000 results. | PubMed Help, read 4 October 2026 |
| Scopus | TITLE-ABS-KEY( ), covering title, abstract and keywords. | "loose phrase" in straight double quotes (curly quotes can cause a parsing error); {exact phrase} in braces, where * is a literal character. | * and ?. No double truncation (*daylight* fails). Plurals and spelling variants are found automatically (anesthesia finds anaesthesia). | W/n (any order) and PRE/n (in order), with words or phrases only. | The help page (updated 24 August 2026) lists OR, then W/n and PRE/n, then AND, then AND NOT. Elsevier announced a change to AND NOT, then AND, then OR, to be complete in early 2026. | No explicit limit on query length. Do not put quotes around single words. | Scopus: Advanced search; Scopus: complex searches; Scopus blog, read 4 October 2026 |
| Web of Science | TS=( ) (Topic: title, abstract, author keywords and Keywords Plus), in Advanced Search, not the default Smart Search. | "phrase". Quotes turn off lemmatization and synonyms ("mouse" does not find mice). | * needs at least 3 letters before it in Topic and Title. ? is one letter, $ is zero or one (odo$r). Wildcards turn off lemmatization, so add both spellings (color* OR colour*). | NEAR/x (NEAR alone means 15 words). No AND inside a NEAR query. | NEAR/x, SAME, NOT, AND, OR. | All Fields searches are capped at 49 operators on one help page and 100 on another, so use TS=. | Web of Science: Search Operators; Web of Science: Search Rules; Web of Science: field tags; Web of Science: Smart Search, read 4 October 2026 |
| EBSCOhost (CINAHL, APA PsycInfo, ERIC) | TI ( ) OR AB ( ). | "phrase"; a wildcard inside quotes still expands. | * expands to at most 2,000 forms and can never be the first character. # is zero or one letter (colo#r). ? is exactly one. Truncation skips alternate spellings (pediatric* misses paediatric). | Nn (any order) and Wn (in order), with n at most 255. | The general help says AND and NOT run before OR; the CINAHL and APA PsycInfo help says left to right. | Field codes in capitals, search words in lower case. Subject headings are MH (CINAHL Headings) and DE (APA Thesaurus), not MeSH. | EBSCO: wildcards; EBSCO: proximity; EBSCO: Booleans; EBSCO: field codes; EBSCO: CINAHL help; EBSCO: APA PsycInfo help, read 4 October 2026 |
| ProQuest | ABSTRACT,TITLE( ), or NOFT( ) for anywhere except the full text. | "phrase". | * is documented as replacing up to five characters, with [*n] for more (nutr[*5]). ? matches zero or one character. No leading wildcard. | NEAR/n (NEAR alone means NEAR/4) and PRE/n. | PRE, NEAR, AND, OR, NOT. | Operators work in lower or upper case. | ProQuest Search Tips; ProQuest field codes, read 4 October 2026 |
| Google Scholar | No field for abstracts; allintitle: or the advanced form's 'in the title of the article'. | Quotes work. | Not documented by Google; failed in 2020 tests. | None. | OR, AND and brackets are not documented and failed in 2020 tests. | Shows at most 1,000 results (Google's help). Accepts about 256 characters (Gusenbauer and Haddaway, 2020). | Google Scholar help; Gusenbauer and Haddaway, 2020, read 4 October 2026 |
| OpenAlex (used for the builder's preview) | title_abstract_keywords. | "phrase" (stemmed unless you ask for exact search). | * needs 3 letters before it and exact search only. The builder sends your full words instead. | "a b"~N. | AND, OR and NOT in capitals. | The whole request URL is limited to about 4 KB. | OpenAlex help, read 4 October 2026 |
Why every concept is bracketed
The databases disagree on order. PubMed reads left to right. Scopus's help page lists OR first, then W/n and PRE/n, then AND, then AND NOT, while Elsevier announced a new order (AND NOT, then AND, then OR) to be complete in early 2026. Web of Science runs NEAR, SAME, NOT, AND, OR. ProQuest runs PRE, NEAR, AND, OR, NOT. EBSCO's general help says AND and NOT run before OR, while its CINAHL and APA PsycInfo help says left to right. Brackets make the order irrelevant, so the builder brackets every concept in every string.
Subject headings: MeSH and the other thesauri
MeSH (Medical Subject Headings) is the US National Library of Medicine's vocabulary for PubMed and MEDLINE. Indexers tag each record with headings, so a heading finds papers that never use your word. In PubMed, [mh] also searches the narrower headings under it unless you write [mh:noexp].
Other databases use their own vocabularies: CINAHL Subject Headings in CINAHL, the APA Thesaurus in APA PsycInfo and the ERIC Thesaurus in ERIC. A MeSH heading pasted into those finds little or nothing, so the builder puts MeSH only in the PubMed string and points you to each database's own thesaurus (EBSCO's help for CINAHL and APA PsycInfo).
How the builder checks a heading
NCBI's automatic term mapping links your words to headings: “remote work” leads to Teleworking, and “HbA1c” to Glycated Hemoglobin. Only a mapping of the whole term counts. “Sense of belonging” maps only its first word, to Sensation, so that mapping is dropped. Each heading left is then looked up by its exact name in NLM's MeSH vocabulary for the current year, and is shown only if it is there, with its D-number and a link to the MeSH Browser. When nothing is found, the builder says “We could not find a MeSH heading for these words”, and your keywords carry the search.
Checking matters. In a 2023 study, 55% of the MeSH terms ChatGPT generated with the best-performing prompt for systematic review searches were not in the MeSH vocabulary (Wang and colleagues, SIGIR 2023).
Recall first
Recall, also called sensitivity, is the share of the relevant papers a search finds. Lagisz and colleagues write that search sensitivity “is calculated as a proportion (or percentage) of indexed benchmark studies (bibliographic records) that are found by a search string” (Research Synthesis Methods, 2025). Precision is the share of what a search finds that is relevant. For a literature review, missing a relevant paper costs more than screening an extra one. The Cochrane Handbook puts it this way: “Searches should aim for high sensitivity, which may result in relatively low precision.”
General chatbots tend to make the opposite trade. Wang and colleagues found that “ChatGPT is capable of generating queries that lead to high search precision, although trading-off this for recall” (SIGIR 2023). So the builder only adds words to yours, keeps outcome and comparison blocks off by default, never uses NOT, and shows a heading only after NLM confirms it.
Check your own string
Pick three to five papers you already know belong in your review. Run the string and see whether each one turns up. A missing paper usually means a concept is short of words its authors used. This is the benchmark check Lagisz and colleagues describe. Our guide to writing a search strategy walks through piloting.
Two worked examples
Example 1: exercise and type 2 diabetes (PICO, with MeSH)
The question: ‘In adults with type 2 diabetes, does a structured exercise programme lower HbA1c compared with usual care?’
Framework: PICO.
- Population (P):
- adults with type 2 diabetes
- Intervention (I):
- a structured exercise programme
- Comparison (C):
- usual care
- Outcome (O):
- HbA1c
Population (P): searched
type 2 diabetes, type II diabetes, T2DM, NIDDM, non-insulin-dependent diabetes, non insulin dependent diabetes, Maturity-Onset Diabetes Mellitus, Maturity-Onset Diabetes, Noninsulin-Dependent Diabetes Mellitus, Type 2 Diabetes Mellitus
Population (P), second concept: off
Off by default: a second concept in the same part narrows the search. Switch it on if every paper must mention it.
adults, adult
MeSH: Adult · D000328
Intervention (I): searched
structured exercise programme, exercise, physical activity, resistance training, aerobic training, exercise training, structured exercise program, structured exercise programmes, physical activities, Active Breaks, Activity Breaks, Acute Exercise, Aerobic Exercise, Isometric Exercise
Comparison (C): off
Off by default: comparisons are often missing from titles and abstracts (Cochrane Handbook 4.4.2).
usual care, standard care
We could not find a MeSH heading for these words.
Outcome (O): off
Off by default: outcomes are often missing from titles and abstracts (Cochrane Handbook 4.4.2; Frandsen and colleagues, 2020).
HbA1c, glycated haemoglobin, glycaemic control, glycated hemoglobin, glycated haemoglobins, glycemic control, glycaemic controls, Fructated Hemoglobins, Glycated Hemoglobin A, Glycated Hemoglobin A1c, Glycated Hemoglobins, Glycohemoglobin
MeSH: Glycated Hemoglobin · D006442, Glycemic Control · D000085002
PubMed
("Diabetes Mellitus, Type 2"[mh] OR "type 2 diabetes"[tiab] OR "type II diabetes"[tiab] OR T2DM[tiab] OR NIDDM[tiab] OR "non-insulin-dependent diabetes"[tiab] OR "non insulin dependent diabetes"[tiab] OR "Maturity-Onset Diabetes Mellitus"[tiab] OR "Maturity-Onset Diabetes"[tiab] OR "Noninsulin-Dependent Diabetes Mellitus"[tiab] OR "Type 2 Diabetes Mellitus"[tiab]) AND ("Exercise"[mh] OR "Resistance Training"[mh] OR "structured exercise programme"[tiab] OR exercis*[tiab] OR "physical activity"[tiab] OR "resistance training"[tiab] OR "aerobic training"[tiab] OR "exercise training"[tiab] OR "structured exercise program"[tiab] OR "structured exercise programmes"[tiab] OR "physical activities"[tiab] OR "Active Breaks"[tiab] OR "Activity Breaks"[tiab] OR "Acute Exercise"[tiab] OR "Aerobic Exercise"[tiab] OR "Isometric Exercise"[tiab])Scopus
TITLE-ABS-KEY("type 2 diabetes" OR "type II diabetes" OR T2DM OR NIDDM OR "non-insulin-dependent diabetes" OR "non insulin dependent diabetes" OR "Maturity-Onset Diabetes Mellitus" OR "Maturity-Onset Diabetes" OR "Noninsulin-Dependent Diabetes Mellitus" OR "Type 2 Diabetes Mellitus") AND TITLE-ABS-KEY("structured exercise programme" OR exercis* OR "physical activity" OR "resistance training" OR "aerobic training" OR "exercise training" OR "structured exercise program" OR "structured exercise programmes" OR "physical activities" OR "Active Breaks" OR "Activity Breaks" OR "Acute Exercise" OR "Aerobic Exercise" OR "Isometric Exercise")PubMed: 21,650 records. OpenAlex: 52,867 works, 1,456 of them reviews. Counted 6 October 2026.
The two counts differ because PubMed and OpenAlex index different journals and treat words differently.
Example 2: first-generation students and belonging (SPIDER, outside health)
The question: ‘How do first-generation university students experience a sense of belonging?’
Framework: SPIDER.
- Sample (S):
- first-generation university students
- Phenomenon of Interest (PI):
- a sense of belonging
- Design (D):
- not named
- Evaluation (E):
- experience
- Research type (R):
- not named
Sample (S): searched
first-generation university students, first-generation students, first-generation college students, first-in-family students, first generation university students, first generation students, first generation college students, first in family students
We could not find a MeSH heading for these words.
Phenomenon of Interest (PI): searched
sense of belonging, belonging, belongingness, connectedness, sense of community, sense of communities, Group Cohesion, Group Cohesiveness, Group Solidarity, Shared Commitment, Shared Community
Evaluation (E): off
Off by default, to keep the search wide. Switch it on to narrow.
experience, experiences, perceptions, views
MeSH: Perception · D010465 (unticked)
The sample has no MeSH heading: no word in its block maps to one as a whole term. “Sense of belonging” has none either, because NCBI maps only its first word, to Sensation. One suggested phrase, “sense of community”, maps to Social Cohesion, which NLM confirms, so that is the only heading in the PubMed string. MeSH is a medical vocabulary, and an education question rests on its keywords.
The Evaluation block shows a trap. “Perceptions” maps to Perception, which NLM confirms, but Perception is MeSH's heading for the senses (its one entry term is Sensory Processing), not for students' views. A confirmed heading exists; it does not always fit. The example leaves it unticked and removes its entry term, as you should for any heading that does not mean what your question means.
EBSCO
(TI ("first-generation university students" OR "first-generation students" OR "first-generation college students" OR "first-in-family students" OR "first generation university students" OR "first generation students" OR "first generation college students" OR "first in family students") OR AB ("first-generation university students" OR "first-generation students" OR "first-generation college students" OR "first-in-family students" OR "first generation university students" OR "first generation students" OR "first generation college students" OR "first in family students")) AND (TI ("sense of belonging" OR belong* OR belongingness* OR connectedness* OR "sense of community" OR "sense of communities" OR "group cohesion" OR "group cohesiveness" OR "group solidarity" OR "shared commitment" OR "shared community") OR AB ("sense of belonging" OR belong* OR belongingness* OR connectedness* OR "sense of community" OR "sense of communities" OR "group cohesion" OR "group cohesiveness" OR "group solidarity" OR "shared commitment" OR "shared community"))Use it in ERIC or APA PsycInfo, and add their own thesaurus terms.
PubMed
("first-generation university students"[tiab] OR "first-generation students"[tiab] OR "first-generation college students"[tiab] OR "first-in-family students"[tiab:~0] OR "first generation university students"[tiab] OR "first generation students"[tiab] OR "first generation college students"[tiab] OR "first in family students"[tiab:~0]) AND ("Social Cohesion"[mh] OR "sense of belonging"[tiab] OR belong*[tiab] OR belongingness*[tiab] OR connectedness*[tiab] OR "sense of community"[tiab] OR "sense of communities"[tiab:~0] OR "Group Cohesion"[tiab] OR "Group Cohesiveness"[tiab] OR "Group Solidarity"[tiab] OR "Shared Commitment"[tiab] OR "Shared Community"[tiab])PubMed: 49 records. OpenAlex: 1,315 works, 2 of them reviews. Counted 6 October 2026.
PubMed, a biomedical database, finds 49 records where OpenAlex finds 1,315. Choose databases for your field, not the familiar ones.
Search strategy table for your methods chapter
PRISMA-S, the reporting guideline for literature searches (Rethlefsen and colleagues, Systematic Reviews, 2021), has 16 items. For each database it asks you to:
- “Name each individual database searched, stating the platform for each” (item 1);
- give each full search strategy “copied and pasted exactly as run” (item 8);
- describe any limits and the reason for them (item 9);
- “provide the date when the last search occurred” (item 13);
- report “the total number of records identified from each database” (item 15).
| Database | Platform | Date searched | Search string as run | Limits | Records found |
|---|---|---|---|---|---|
| PubMed | PubMed | 6 October 2026 | ("Diabetes Mellitus, Type 2"[mh] OR "type 2 diabetes"[tiab] OR "type II diabetes"[tiab] OR T2DM[tiab] OR NIDDM[tiab] OR "non-insulin-dependent diabetes"[tiab] OR "non insulin dependent diabetes"[tiab] OR "Maturity-Onset Diabetes Mellitus"[tiab] OR "Maturity-Onset Diabetes"[tiab] OR "Noninsulin-Dependent Diabetes Mellitus"[tiab] OR "Type 2 Diabetes Mellitus"[tiab]) AND ("Exercise"[mh] OR "Resistance Training"[mh] OR "structured exercise programme"[tiab] OR exercis*[tiab] OR "physical activity"[tiab] OR "resistance training"[tiab] OR "aerobic training"[tiab] OR "exercise training"[tiab] OR "structured exercise program"[tiab] OR "structured exercise programmes"[tiab] OR "physical activities"[tiab] OR "Active Breaks"[tiab] OR "Activity Breaks"[tiab] OR "Acute Exercise"[tiab] OR "Aerobic Exercise"[tiab] OR "Isometric Exercise"[tiab]) | None | 21,650 |
| Scopus | Elsevier | The day you run it | TITLE-ABS-KEY("type 2 diabetes" OR "type II diabetes" OR T2DM OR NIDDM OR "non-insulin-dependent diabetes" OR "non insulin dependent diabetes" OR "Maturity-Onset Diabetes Mellitus" OR "Maturity-Onset Diabetes" OR "Noninsulin-Dependent Diabetes Mellitus" OR "Type 2 Diabetes Mellitus") AND TITLE-ABS-KEY("structured exercise programme" OR exercis* OR "physical activity" OR "resistance training" OR "aerobic training" OR "exercise training" OR "structured exercise program" OR "structured exercise programmes" OR "physical activities" OR "Active Breaks" OR "Activity Breaks" OR "Acute Exercise" OR "Aerobic Exercise" OR "Isometric Exercise") | None | The count Scopus shows |
The builder fills this table in from your blocks. Tick the databases you ran, paste in any string you changed, and type the counts for databases we cannot count for you.
Download a blank template: Word (.docx) or CSV for Excel or Google Sheets.
The records found feed the identification box of a PRISMA flow diagram.
What the builder cannot do
- The strings are a starting point: pilot them, and show your strategy to a subject librarian if you can.
- Translating syntax does not translate subject headings: MeSH works in PubMed and MEDLINE only.
- A confirmed heading can still be the wrong sense of your word: “perceptions” maps to Perception, MeSH's heading for the senses. Untick any heading that does not fit, and remove its entry terms.
- The OpenAlex count is OpenAlex's own. It runs your full words with its own stemming, not your asterisks, in titles, abstracts and keywords. Other databases cover different journals, so their counts will differ. Only the PubMed count comes from PubMed.
- Google Scholar is a supplementary search: it shows at most 1,000 results and, in published tests, handled Boolean strings poorly.
- It does not write proximity searches (NEAR, W/n), study-design filters, date or language limits, or Embase and Ovid syntax.
- ProQuest documents its * as standing for up to five more letters, so add any longer word you need as its own term.
- Database searching is one route. Citation chaining and grey literature find what databases miss.
- The AI suggestions can be off topic. Check every word before you run the string.
- Database rules change. Every rule on this page links its source and the date we checked it.
Other free tools for searching
- MeSH Browser (NLM): look up any heading, its entry terms and its place in the tree.
- MeSH on Demand (NLM): paste an abstract to get suggested MeSH terms. NLM says its suggestions are machine-generated and do not reflect human review.
- Polyglot Search Translator (Bond University): translates a finished PubMed or Ovid MEDLINE string into other databases' syntax. A 2022 review notes that complicated searches with wildcards may not translate accurately (Kung, JCHLA, 2022).
Questions
How do I create a search string for a literature review?
Split your question into its main concepts, usually who it is about and what is being done or studied.
For each concept, list the words authors use for it: synonyms, abbreviations, British and American spellings, and word endings. Join the words inside each concept with OR, put brackets around each concept, and join the concepts with AND. Then add the database's subject headings, such as MeSH in PubMed, to the matching concept.
The builder above drafts each step, and you can edit every word. Then pilot the string against three to five papers you already know belong in your review.
What does a search string look like?
For ‘Does exercise lower HbA1c in adults with type 2 diabetes?’, a PubMed string could be:
("Diabetes Mellitus, Type 2"[mh] OR "type 2 diabetes"[tiab] OR T2DM[tiab]) AND ("Exercise"[mh] OR exercis*[tiab] OR "physical activit*"[tiab])Each bracket is one concept. In Scopus the same search sits inside TITLE-ABS-KEY( ) with no MeSH headings, because MeSH is PubMed's vocabulary.
How many keywords should a literature review search have?
There is no fixed number. What matters is that each concept has enough alternatives to catch papers that describe it differently.
The concepts themselves should be few. The Cochrane Handbook says search strategies should avoid using too many different search concepts, but should combine a wide variety of search terms with OR within each concept. Every extra AND block narrows the search.
What is the difference between a search strategy and a search string?
The string is the line you paste into one database.
The strategy is the whole plan: which databases you searched, the string you ran in each, the limits you set, the date you ran them, how many records each returned, and the other ways you found papers, such as checking reference lists. PRISMA-S lists 16 items to report, and the table at the end of the builder holds the ones that belong to each database search.
Can I use the same search string in every database?
No. Each database has its own field codes, truncation rules and order of operations, and subject headings differ between them: MeSH in PubMed, CINAHL Subject Headings in CINAHL, the APA Thesaurus in APA PsycInfo.
The builder writes a separate string for each database from the same blocks, and the syntax table above shows where they differ.
Does Google Scholar support truncation, brackets or OR?
Google's help for Scholar does not document truncation, brackets or OR.
In tests published in 2020, Gusenbauer and Haddaway found that Scholar failed their checks of OR, AND, brackets and truncation, and that it accepts searches of up to 256 characters. So the builder gives you several short Scholar searches instead of one long string.
Scholar's own help says it shows at most 1,000 results for any search, so treat it as a supplementary source.
Should I use NOT in my search string?
Rarely. NOT removes every record that mentions the excluded word, including relevant papers that mention it in passing. The Cochrane Handbook says the NOT operator should be avoided where possible.
Exclude studies at screening instead, where you can see why each one is out. The builder never adds NOT.
Do I need MeSH terms if my topic is not health?
MeSH is the vocabulary of PubMed and MEDLINE, so it only helps if you search them. Education, psychology and social science databases have their own, such as the ERIC Thesaurus and the APA Thesaurus.
The builder checks MeSH for every question, says plainly when it found no heading, and keeps your keywords as the main search.
How do I find the right MeSH term?
Look your word up in NLM's MeSH Browser. If it is an entry term, MeSH sends you to the heading: ‘remote work’, for example, leads to Teleworking.
The builder does this for each concept and shows only headings that NLM's vocabulary confirms, each with its ID and a link to its MeSH record. Add a word to any block and press Update results to check it.
Can ChatGPT write my search string?
A general chatbot can write a string that looks right.
In a 2023 study, Wang and colleagues found that ChatGPT's Boolean queries for systematic reviews gained precision but lost recall. With the best-performing prompt, 55% of the MeSH terms it generated were not in the MeSH vocabulary. For a literature review, missing relevant papers is the costlier mistake.
The builder uses an AI model only to suggest extra words, keeps all of yours, and shows a MeSH heading only after the National Library of Medicine's vocabulary confirms it.
How do I report my search in the methods chapter?
Name each database and the platform you searched it on. Give the full string for each exactly as you ran it, state the limits you used and why, give the date of the last search, and report how many records each database returned.
These are items 1, 8, 9, 13 and 15 of PRISMA-S (Rethlefsen and colleagues, Systematic Reviews, 2021). The table at the end of the builder lays them out for you to copy into your chapter or appendix.
Is the builder free, and is my question saved?
It is free, with no sign-up.
Your question is not stored in a database or linked to any account. It is sent to an AI model to suggest synonyms, and the words in your blocks are sent to the National Library of Medicine, PubMed and OpenAlex to check headings and count results. To avoid repeating work, our server keeps your question and its suggestions in memory for up to a week, then drops them.
Nothing goes into a Litrevu library unless you press the button to add papers. See our privacy policy for the AI providers we use.
Sources
- PubMed Help, National Library of Medicine, read 4 October 2026.
- How can I best use the Advanced search? Scopus Support Center (updated 24 August 2026), read 4 October 2026.
- Search Tips: What makes a Scopus search complex? Scopus Support Center (updated 20 October 2025), read 4 October 2026.
- Boolean searches in Scopus: understanding operator precedence and best practices, Scopus blog (24 March 2025), read 4 October 2026.
- Search Operators, Web of Science help (updated 27 August 2026), read 4 October 2026.
- Search Rules, Web of Science help (23 July 2026), read 4 October 2026.
- Web of Science Core Collection Advanced Search Field Tags, Web of Science help (updated 27 August 2026), read 4 October 2026.
- Smart Search, Web of Science help (16 September 2026), read 4 October 2026.
- Searching with Wildcards and Truncation Symbols, EBSCOhost help, read 4 October 2026.
- Proximity Searches, EBSCOhost help, read 4 October 2026.
- Booleans, EBSCOhost help, read 4 October 2026.
- Field Codes, EBSCOhost help, read 4 October 2026.
- CINAHL Plus with Full Text database help, EBSCOhost, read 4 October 2026.
- APA PsycInfo database help, EBSCOhost, read 4 October 2026.
- Search Tips, ProQuest help, read 4 October 2026.
- Field codes, ProQuest help, read 4 October 2026.
- Google Scholar Help, read 4 October 2026.
- Gusenbauer M, Haddaway NR. Which academic search systems are suitable for systematic reviews or meta-analyses? Evaluating retrieval qualities of Google Scholar, PubMed, and 26 other resources. Research Synthesis Methods 2020;11(2):181-217, read 4 October 2026.
- Searching, OpenAlex help (updated 3 October 2026), read 4 October 2026.
- Lefebvre C, Glanville J, Briscoe S, et al. Chapter 4: Searching for and selecting studies (last updated March 2025). Cochrane Handbook for Systematic Reviews of Interventions, version 6.5.1, read 4 October 2026.
- Frandsen TF, Bruun Nielsen MF, Lindhardt CL, Eriksen MB. Using the full PICO model as a search tool for systematic reviews resulted in lower recall for some PICO elements. Journal of Clinical Epidemiology 2020;127:69-75, read 4 October 2026.
- Methley AM, Campbell S, Chew-Graham C, McNally R, Cheraghi-Sohi S. PICO, PICOS and SPIDER: a comparison study of specificity and sensitivity in three search tools for qualitative systematic reviews. BMC Health Services Research 2014;14:579, read 4 October 2026.
- Cooke A, Smith D, Booth A. Beyond PICO: the SPIDER tool for qualitative evidence synthesis. Qualitative Health Research 2012;22(10):1435-1443, read 4 October 2026.
- Moola S, Munn Z, Sears K, et al. Conducting systematic reviews of association (etiology): The Joanna Briggs Institute's approach. International Journal of Evidence-Based Healthcare 2015;13(3):163-169, read 4 October 2026.
- Peters MDJ, Marnie C, Tricco AC, et al. Updated methodological guidance for the conduct of scoping reviews. JBI Evidence Synthesis 2020;18(10):2119-2126, read 4 October 2026.
- Aslam S, Emmanuel P. Formulating a researchable question: A critical step for facilitating good clinical research. Indian Journal of Sexually Transmitted Diseases and AIDS 2010;31(1):47-50, read 4 October 2026.
- Wang S, Scells H, Koopman B, Zuccon G. Can ChatGPT Write a Good Boolean Query for Systematic Review Literature Search? SIGIR 2023, pp. 1426-1436 (arXiv 2302.03495), read 4 October 2026.
- Lagisz M, Yang Y, Young S, Nakagawa S. A practical guide to evaluating sensitivity of literature search strings for systematic reviews using relative recall. Research Synthesis Methods 2025;16(1):1-14, read 4 October 2026.
- Rethlefsen ML, Kirtley S, Waffenschmidt S, et al. PRISMA-S: an extension to the PRISMA Statement for Reporting Literature Searches in Systematic Reviews. Systematic Reviews 2021;10:39, read 4 October 2026.
- MeSH RDF lookup service, US National Library of Medicine, and its acceptable use policy, read 4 October 2026.
- MeSH on Demand, US National Library of Medicine, read 4 October 2026.
- Kung J. Polyglot Search Translator. Journal of the Canadian Health Libraries Association 2022;43(1), read 4 October 2026.
Related guides
- How to Write a Search Strategy for a Literature Review
- How to Find Academic Sources for Your Research Paper (2026 Guide)
- How to Write a Research Question (Step-by-Step, With Examples)
- Snowballing in a Literature Review: How to Find Papers by Citation Chaining (2026)
- How to Find and Use Grey Literature in a Literature Review (2026)
- How to Write Inclusion and Exclusion Criteria (With Examples)
- Scoping vs Systematic vs Narrative Review: Which Do You Need?
- How to Screen 500 Abstracts Without Reading Every One (2026)
- PRISMA Flow Diagram: How to Report Your Search
Other free tools
- Dissertation Word Count Calculator
How long your literature review and every other chapter should be for your word limit and level, how many sources to cite, and a dated plan for each stage up to your hand-in date, from published university guidance.
- Reference Checker
Paste a reference list to see which references are real, which details are wrong, and which papers are retracted, checked against Crossref and OpenAlex.
- AI Disclosure Statement Generator
Pick your university and what you used AI for to get its own AI declaration filled in, a reference for the tool in APA, Harvard or MLA, and the records to keep.
- Literature Review Matrix
Drop in up to three of your own PDFs to get a comparison table where each filled cell quotes the passage it came from, or download a blank matrix template in Excel, Word or CSV.
For librarians: every database rule here links its source and date, and the builder needs no account. See Litrevu for universities.