Forty tabs, no plan
You search a bit everywhere, find something, lose it again, and can't say later how you found what you found.
A prompt for ChatGPT and Claude · for theses and literature reviews
A 7-phase prompt that takes you from a vague topic to tested search strings, a tidy Zotero library and a reference list in APA 7. Every source comes with a DOI, PMID or ISBN, or it gets flagged.
One-time purchase on PromptBase · versions for ChatGPT and Claude
You search a bit everywhere, find something, lose it again, and can't say later how you found what you found.
Chatbots can produce citations that look real but lead nowhere. One fake source in your reference list is one too many.
PubMed, Scopus, PsycINFO, Google Scholar: each has its own syntax. Synonyms, MeSH terms and Boolean operators are easy to get wrong.
Paste the prompt, fill in your topic, field and level. The AI works through one phase per answer, ends with a checklist and waits until you are ready for the next one.
A searchable research question (PICO or another fitting framework), a concept table with synonyms and thesaurus terms, include/exclude criteria.
VPN, institutional login or library proxy, explained for your system. No university? Legal open-access routes instead.
Install Zotero and the browser connector, with exact menu paths for metadata, PDF retrieval and your citation style.
Database recommendations for your field and a ready-to-paste search string for each one. With web search on, strings for open databases are tested with real hit counts.
A collection and tag scheme for Zotero, then step-by-step import and screening against your criteria.
Zotero's Word plugin, your first citation and a reference list in APA 7 or any style you choose.
A final checklist and a short summary of your search for the methods chapter.
Two excerpts, shortened but not edited. Nothing here was written for this page.1
| Concept | Free-text synonyms | MeSH (PubMed) | ERIC descriptor |
|---|---|---|---|
| Mindfulness (I) | mindfulness, mindful*, MBSR, MBCT, "mindfulness-based stress reduction", meditation, "body scan" […] | Mindfulness D064866 checked Mindfulness-Based Stress Reduction D000099024 checked Meditation D019122 checked |
"Mindfulness" is not a descriptor: ERIC says to use Metacognition instead checked |
| Exam anxiety (O) | "test anxiety", "exam anxiety", "examination stress", "test-taking anxiety", "evaluation anxiety" […] | Test Anxiety D000083563 checked broader: Performance Anxiety D060805 checked |
Test Anxiety checked |
Vocabulary pitfall from the same answer
"Test Anxiety" in MeSH only exists since 2021. Before that, NLM indexed these papers under the general term "Anxiety" (1963–2020). Papers from 2016–2020 therefore won't carry the Test Anxiety heading. Free-text terms ([tiab]) are essential.
("Diabetes Mellitus, Type 2"[mh] OR "type 2 diabet*"[tiab] OR "type II diabet*"[tiab] OR T2DM[tiab] OR T2D[tiab] OR NIDDM[tiab] OR "non-insulin-dependent diabet*"[tiab] OR "noninsulin dependent diabet*"[tiab] OR "adult-onset diabet*"[tiab])
AND
("High-Intensity Interval Training"[mh] OR "high-intensity interval*"[tiab] OR "high intensity interval*"[tiab] OR HIIT[tiab] OR "interval training"[tiab] OR "interval exercise*"[tiab] OR "intermittent exercise*"[tiab] OR "high-intensity intermittent*"[tiab] OR "sprint interval*"[tiab] OR "aerobic interval*"[tiab] OR REHIT[tiab])
NOT (animals[mh] NOT humans[mh])
NOT (mice[ti] OR mouse[ti] OR rat[ti] OR rats[ti] OR murine[ti])
AND 2015:3000[dp]
AND (english[la] OR german[la])
The first version used an RCT filter and missed a relevant trial whose abstract says neither "random*" nor "trial". The AI noticed it in the known-item test, removed the filter and said why.
We asked ChatGPT: "Find sources for my bachelor's thesis on productivity effects of remote and hybrid work." The sources it gave were real. What was missing is the part your methods chapter needs.2
| One-line prompt | Literature Review Assistant | |
|---|---|---|
| Result | A list of sources to start with | A documented search you can repeat and report |
| Research question | Taken as typed | Sharpened with PICO or a fitting framework |
| Thesaurus terms | None | MeSH, ERIC, APA Thesaurus, marked "checked" or "(verify)" |
| Search strings | None | One per database, in its own syntax |
| Testing | None | Hit counts and a known-item test in open databases |
| Source rule | Not stated | DOI, PMID, ISBN or URL, or flagged "UNVERIFIED" |
| Next steps | Up to you | Zotero setup, screening, citing in Word |
One page with Boolean operators, truncation, phrase search and field tags for PubMed, Scopus, Web of Science, EBSCO and Google Scholar. Print it, pin it next to your screen.
That depends on your institution's rules on AI use. The prompt reminds you to check them and to disclose AI use where required. It never queries licensed databases for you and never downloads paywalled PDFs.
Yes. Enter "none" as your university and the prompt switches to open databases and legal routes such as open access, public libraries and interlibrary loan.
APA 7th Edition by default. You can set any other style, for example Harvard, Chicago, MLA or IEEE, and Zotero formats your reference list in it.
Buy the version for the chatbot you use. Both follow the same 7 phases and rules. Turn on web search so the AI can check thesaurus terms and test search strings.
No. Phase 2 walks you through installing and setting up Zotero with exact menu paths.
No. You search licensed databases yourself. The prompt never suggests shadow libraries and points you to legal ways in: open-access versions, interlibrary loan or emailing the author.
No. It plans and documents your literature search. Reading, judging and writing stay with you.
Purchases run through PromptBase, so PromptBase's terms apply.
Seven phases, tested strings, sources you can check.