thinkhard.SI

Prompt repository Research & learning

Write search strings for a literature or database search

You need keyword and Boolean searches for Google Scholar, a library database or a news archive that find the right sources without thousands of misses.

Registry no.
RES-021
Collection
Research & learning
Laid out for
Any chat model
Score
10 out of 10
Blanks to fill
6
Added

Research & learningAny chat

The prompt

Score 10 of 10, Excellent

Use it when You need keyword and Boolean searches for Google Scholar, a library database or a news archive that find the right sources without thousands of misses.

  • {{research_question}}
  • {{purpose}}
  • {{databases}}
  • {{inclusion}}
  • {{exclusion}}
  • {{known_sources}}
Full prompt and why it works
Prompt
Research question: {{research_question}}
What the search is for: {{purpose}}
Where I will search: {{databases}}
Include only (years, languages, countries, types of source): {{inclusion}}
Leave out: {{exclusion}}
Two or three sources I already know are relevant: {{known_sources}}

Task:
Write search strings I can paste into the databases above to find sources for my question. A good search is one that finds the known relevant sources above and keeps the results to a number I can screen, usually a few hundred at most.

Format:
1. Concepts: split the question into 2 to 4 main concepts. For each, list synonyms, related terms, US and UK spellings, and acronyms, in a small table.
2. Search strings: one per database, in a code block, using that database's own syntax for AND, OR, phrase quotes, truncation and field limits such as title or abstract. Where a database works differently, say so; Google Scholar, for instance, ignores some operators and limits how long a search can be.
3. Filters to set in each database's menus, matching my include and leave-out lists.
4. How to test it: check that the known sources come up, and if not, which term to add. If results run to thousands, which concept to tighten first; if only a handful, which to loosen.
5. A search log line for each database (database, date, string, filters, number of results) for me to fill in, since reviews often have to report how the search was done.

Requirements:
- Use only syntax you are confident each database supports. If you are unsure about a feature in a named database, say so and give a simpler version that works everywhere.
- Do not suggest specific papers, authors or journals from memory, because the search itself should find the sources.
- Under 400 words, plus the search strings.

Example of one string's form only, on an invented question: ("remote work" OR telework* OR "work from home") AND (onboarding OR "new hire*") AND (retention OR turnover)

If my question is too broad to search well, say so and suggest how to narrow it before searching.

Variables

{{research_question}}
Your question, as specifically as you can put it.
{{purpose}}
For example "a literature review chapter" or "background for a policy briefing".
{{databases}}
For example "Google Scholar, PubMed and our library's EBSCO search".
{{inclusion}}
For example "2015 onward, English, peer-reviewed studies and government reports".
{{exclusion}}
For example "opinion pieces, studies of children". Write "none" if nothing.
{{known_sources}}
Titles or short notes on two or three sources you know belong in the results.

Why it works

  • An example. The sources you already know act as a test: if the search does not find them, you are told which term to add.
  • Format. A concept table of synonyms and variants, then one string per database in its own syntax, makes the search repeatable and easy to adjust.
  • If unsure. Where it is unsure a database supports a feature, it says so and gives a simpler string that works everywhere, instead of syntax that silently fails.

How to use it

  1. Press Fill in variables and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/literature-search-strings

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "literature-search-strings"}, or just ask for it by name.