home/productivity/deep-research-prompt

Deep Research Prompt

GPTClaudeDeepSeek··149 copies·updated 2026-07-14
deep-research-prompt.prompt
# Stage 1 — Deep Research Prompt

> **How to use this template**
> 1. Copy this file to a working location — **once per chapter** of your outline.
> 2. Replace every `{{PLACEHOLDER}}`. The CONTEXT block stays the same across all runs; the TASK and SCOPE blocks change each run.
> 3. Delete every bracketed `[guidance note]`.
> 4. Run each filled prompt in **Google Gemini with Deep Research mode enabled**, as a separate conversation.
> 5. Save each result to its own document, named by chapter.
>
> Run one chapter per conversation — not the whole work at once. Deep Research has a finite attention budget per task; focused runs return deeper, better-sourced results.

---

## CONTEXT — read carefully, this frames the task

I am writing a {{WORK_TYPE}} in the field of {{FIELD_OF_STUDY}}.

**Topic:** {{WORK_TOPIC}}

**Overview of the work:** {{WORK_OVERVIEW — 3–6 sentences. The same overview in every run, so each research pass shares context.}}

**Structure of the work:** {{HIGH_LEVEL_OUTLINE — the full chapter list from Stage 0, so the model knows where this chapter sits in the whole.}}

**Already covered in earlier runs:** {{ALREADY_COVERED — name the chapters already researched, so this run does not duplicate them. For the first run, write "none — this is the first research run".}}

**Bibliography requirements for the work:** {{BIBLIOGRAPHY_REQUIREMENTS — total source count, how many must be peer-reviewed (journal articles, monographs, dissertations), and any language requirements.}}

## YOUR TASK IN THIS RUN

Collect authoritative sources and detailed theses for **{{SECTION_TITLE}}**.

I do not need a bare list of links. I need **detailed theses**: for each source, what exactly the author claims, what data or figures they cite, what conclusions they draw, and how it can be used in my work.

**Critical requirements:**

- Every thesis is tied to a specific source — author(s), title, year, journal or publisher, volume/issue and pages where applicable, and a DOI or URL where available.
- Priority of source types: peer-reviewed journal articles, monographs, official reports and standards. Industry analyst reports are acceptable for market data. Avoid blogs and unattributed web pages except as supplementary material.
- Recency: {{RECENCY — e.g. "prioritise 2020–2026, but foundational classic works are also required".}}
- Languages: {{LANGUAGE_MIX — e.g. "roughly 60% English, 40% [language]", or simply the source language(s) to use.}}
- **Do not invent sources.** Do not approximate a citation you are unsure of. If you cannot verify a source's details, say so explicitly rather than guessing. A flagged uncertainty is useful; a fabricated citation is harmful.
- Where a source includes a table, chart, or figure that would be usable in the work, describe it and note where it can be found.

## SCOPE OF THIS RUN

Below are the subsections of this chapter and the specific questions each one needs sources for. Cover every subsection.

[For each subsection, fill in the block below. Duplicate the block as many times as there are subsections.]

### {{SUBSECTION_ID}}. {{SUBSECTION_TITLE}}

Sources and theses needed on:

{{SUBSECTION_QUESTIONS — the specific questions, sub-topics, debates, and data points that need sources. If the field has canonical authors or works that MUST be covered for this subsection, name them here. Be concrete: specific questions yield specific, usable theses; vague prompts yield filler.}}

[— end of subsection block; duplicate as needed —]

## OUTPUT FORMAT

For each subsection, produce:

### [Subsection number and title]

**THESES WITH SOURCES:**

1. [A developed thesis — 3–5 sentences, with concrete data or figures] → **Source:** [Author(s), "Title", Journal/Publisher, Year, Volume/Issue, Pages, DOI/URL]
2. …

(Minimum {{MIN_THESES}} theses per subsection. For chapters central to the work's argument, aim higher.)

**ADDITIONAL SOURCES** (to broaden the bibliography — brief: author, title, year, type):

- …

**USABLE TABLES / FIGURES** (if any source contains data displays worth reproducing):

- [Description of the table or figure, the source, and where in it to find them.]

**GAPS:** If sources are scarce or absent for any question above, state it plainly. Do not fill a gap with a weak or invented source.

fill the variables

This prompt has 8 variables. Pro fills them into a ready-to-paste prompt for you — no manual find-and-replace.

{{PLACEHOLDER}{{WORK_TYPE}{{FIELD_OF_STUDY}{{WORK_TOPIC}{{SECTION_TITLE}{{SUBSECTION_ID}{{SUBSECTION_TITLE}{{MIN_THESES}
Unlock with Pro →

when to use it

Community prompt sourced from the open-source GitHub repo mcqx4/scriptorium (MIT). A "Deep Research Prompt" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

productivitycommunitydeveloper

source

mcqx4/scriptorium · MIT