
My Ultimate DeepResearch Prompt Builder Template and How I Use It
The Exact Prompt Engineering System Powering My DeepResearch Workflow

Table of Contents
The Ultimate DeepResearch Workflow
This blogpost was published on my X/Twitter account on June 20th, 2025 .
TL;DR: I feed the template below into Gemini 2.5 Pro to build the DeepResearch prompt. Then I use the output to run DeepResearch with. You’ll find more context further down, but the main idea is simple: Just drop your core ideas between
TOPIC BEGINS HERE
andTOPIC ENDS HERE
. The rest builds itself.
Google just doesn’t cut it anymore: I’m the guy who wired a mini–data center into his basement. When you’ve got almost 3-dozen GPUs humming at 3 a.m. and a brain that treats half-baked ideas like Pokémon’s you gotta catch ’em all, shallow Googling just doesn’t cut it. I needed a research system that could keep up with the chaos in my head, force clarity, and let me ship faster than my cats can yank the UPS cable while livestreaming (true story ).
DeepResearch and this framework turn my chaotic untangled thoughts into informative, in-depth, comprehensive reports. I also use it to learn anything and I have it wired into the loop of how I code with agents. Today, I’m sharing this workflow with you.
This is blogpost #2 in my 101 Days of Blogging . If it sparks anything; ideas, questions, or critique, my DMs are open. Hope it gives you something useful to walk away with.
From Brainstorm to Battle-Ready Report
I believe that if you genuinely want to move the needle, whether you’re an indie builder, a founder hunting for market clarity, or just someone tired of getting subpar answers, you need a proper system. Something structured that transforms vague curiosities into pinpoint insights, ruthlessly forces clarity, prevents endless rabbit holes, and delivers actual value (think high signal, zero noise). To me, that’s DeepResearch.
Why DeepResearch Exists (And Why You Need a Real Prompt Framework)
Before DeepResearch, I’d “just check one thing” and suddenly I’m 138 tabs deep with outdated blogposts and conflicting info. DeepResearch fixed that, but only because I learned how to use it. There’s a method to the clarity.
DeepResearch, among many things, could be:
- The scientific method, but supercharged by AI for extracting real knowledge.
- A tactical framework built for cutting through noisy markets.
- Your tireless co-thinker who never shortcuts logic or takes naps.
Traditional research often ends up vague, ambiguous, and misses key insights entirely. So I built-and obsessively iterated-a prompt framework designed explicitly to fix these problems. This approach guides both me and the AI to:
- Set crystal-clear goals upfront.
- Ruthlessly define the research scope.
- Clearly define terms to prevent hallucinations.
- Break the research into logical, manageable tasks (think reasoning, not just dumping everything).
- Enforce rigorous evidence, credible sources, and sanity checks.
- Let me fine-tune the AI’s tone: whether I need a sarcastic analyst, an academic, or an executive perspective.
This became the Ultimate DeepResearch Prompt Builder Template, the backbone of every serious AI-driven research I execute.
The Template That Creates Any DeepResearch Prompt
I feed the template below into Gemini 2.5 Pro to build the DeepResearch prompt. Then I use the output to run DeepResearch with. You’ll find more context further down, but the main idea is simple: Just drop your core ideas between TOPIC BEGINS HERE
and TOPIC ENDS HERE
. The rest builds itself.
# **Ultimate Deep Research Prompt Builder Template**
**Instructions:** Fill in the placeholders (`[ ]`) below based on the provided topic to create a highly detailed and effective prompt for generating a comprehensive deep research report. The more specific, clear, and thorough you are in defining each section, the higher the quality and relevance of the AI's output will be.
**Topic:**
TOPIC BEGINS HERE
"""
"""
TOPIC ENDS HERE
---
**You are tasked with generating a deep research report.** Please follow the structure, instructions, and guidelines below meticulously.
## 1. Objective & Core Question(s)
* **Explanation:** Clearly define the primary goal, the central question(s) this research must answer, and the specific, tangible outcome you expect from the report. Precision here is crucial for focusing the research.
* **Instruction:** State the main purpose, the core question(s), and the desired end product of this research. Be specific about what success looks like.
* **Placeholders:**
* **Main Research Topic:** `[Re-state the overarching subject clearly.]`
* **Central Research Question(s):** `[List the specific, core question(s) the research must answer. Be precise and focused. E.g., 'What are the primary drivers of X?', 'How does Y compare to Z based on criteria A, B, C?', 'Is hypothesis P valid under conditions Q?']`
* **Desired Outcome & Report Goal:** `[Specify the tangible outcome. E.g., 'A comprehensive analysis of...', 'A feasibility study for...', 'Strategic recommendations for...', 'A comparative evaluation of...', 'A market landscape overview detailing...', 'A data-driven report validating/invalidating hypothesis X...']`
## 2. Context, Background & Scope
* **Explanation:** Provide all necessary background information, definitions, known data, your current understanding, and any initial hypotheses. Crucially, define the boundaries to prevent scope creep.
* **Instruction:** Include relevant history, current state, key players/factors, existing data points, initial assumptions, and specify the research's limits (timeframe, geography, exclusions).
* **Placeholders:**
* **Relevant Background Information:** `[Provide essential context: history, significance, current situation, key players, existing relevant data points or known facts.]`
* **Initial Hypothesis/Perspective (If Any):** `[State your starting assumption, preliminary thoughts, or the specific angle/lens you want the research to consider. E.g., 'My hypothesis is that [X] is declining due to [Y]...', 'We believe [Technology A] offers significant advantages over [Technology B] because...']`
* **Define Scope and Boundaries:** `[Specify exactly what should be included and EXCLUDED. E.g., 'Focus only on the US market', 'Exclude companies with less than $XM revenue', 'Limit analysis to non-profit sector', 'Do not analyze historical data before 2015'.]`
* **Specify Timeframe:** `[Define the historical period or future projection horizon. E.g., 'Analyze data from the past 5 years (2019-2024)', 'Project trends for the next 10 years', 'Focus on the period Q1 2023 - Q1 2024'.]`
* **Specify Geographic Limits (if any):** `[Define the geographical area. E.g., 'Global analysis', 'Focus on the EU region', 'Compare trends in North America vs. Southeast Asia'.]`
## 3. Key Terms & Definitions
* **Explanation:** Define any specific terminology, acronyms, concepts, or jargon central to the topic to ensure the AI uses them correctly and consistently, avoiding ambiguity.
* **Instruction:** List and clearly define critical terms that are essential for an accurate understanding and execution of the research.
* **Placeholder:** `[Define Key Terms: List and provide precise definitions for essential terms, concepts, or acronyms within the context of this research. Example: 'Define 'Sustainable Growth Rate (SGR)' as used in corporate finance.', 'Define 'Platform Economy' specifically in relation to the gig workforce.']`
## 4. Step-by-Step Research Instructions & Sub-Tasks
* **Explanation:** Break down the main research objective into logical, sequential steps or distinct areas of investigation. Guide the AI through *exactly* what needs to be explored, analyzed, and synthesized.
* **Instruction:** Provide a detailed, itemized list of the required research areas and the specific questions to answer or tasks to complete within each step. Be explicit.
* **Placeholders (Adapt, Add, or Remove Steps as Needed):**
* **Step 4.1: Foundational Analysis:** `[E.g., 'Analyze the core principles/mechanisms of [Subject]', 'Detail the historical evolution of [Topic]', 'Explain the fundamental business model of [Industry/Company]'. Specify key questions to answer within this step.]`
* **Step 4.2: Data Gathering & Trend Analysis:** `[E.g., 'Gather and analyze market size data for [Product/Service] for the specified timeframe [Timeframe]', 'Investigate trends in [Metric A] and [Metric B]', 'Present historical performance data for [Entity/Indicator]'. Specify data sources if known.]`
* **Step 4.3: Competitive/Comparative Landscape:** `[E.g., 'Identify key competitors/alternatives to [Subject]', 'Profile [Competitor A, B, C], analyzing their strengths, weaknesses, strategies, market share', 'Compare [Case Study 1] with [Case Study 2] based on criteria [X, Y, Z]'.]`
* **Step 4.4: Factor Analysis (Drivers, Challenges, Influences):** `[E.g., 'Identify and analyze the key drivers of growth/decline in [Market/Topic]', 'Investigate the impact of [Regulatory Change/Technological Advance]', 'Examine challenges and barriers related to [Subject]', 'Analyze the influence of macroeconomic factors like [Factor A, B]'.]`
* **Step 4.5: Stakeholder/Customer Analysis:** `[E.g., 'Analyze the perspectives and influences of key stakeholder groups ([Group 1], [Group 2])', 'Investigate customer adoption patterns, behaviors, and needs related to [Product/Topic]'.]`
* **Step 4.6: Critical Analysis & Synthesis:** `[Instruct specific synthesis. E.g., 'Apply [SWOT/PESTLE/Porter's Five Forces] framework to synthesize findings related to [Subject]', 'Critically evaluate the strengths, weaknesses, opportunities, and threats based *specifically* on the research findings from previous steps', 'Assess the overall viability/potential/impact of [Subject] synthesizing data from all areas researched'.] **Crucially, synthesize findings – do not just list information.**`
* **Step 4.7: Conclusion & Recommendations (If Applicable):** `[Based on the objective, instruct conclusion type. E.g., 'Formulate a concluding assessment addressing the main research question(s)', 'Provide data-driven, actionable recommendations for [Target Audience/Stakeholder] based *directly* on the synthesized analysis', 'Assess the validity of the initial hypothesis based on the evidence gathered'.]`
* `[Add further specific steps/sub-tasks as required by your objective.]`
## 5. Desired Output Format & Report Structure
* **Explanation:** Specify precisely how the final research output should be organized, structured, and formatted for clarity and usability.
* **Instruction:** Detail the required report structure, sections, formatting guidelines, data visualization needs, and citation style.
* **Placeholders:**
* **Report Structure:** `[Provide a specific list of required sections and their order. E.g., 1. Executive Summary (max 1 page), 2. Introduction (including Objective & Scope), 3. Methodology Overview, 4. Detailed Findings (structured according to Section 4 Steps), 5. Critical Analysis & Synthesis (e.g., SWOT/PESTLE), 6. Conclusion (addressing research questions), 7. Recommendations (if applicable), 8. References, 9. Appendix (if needed).]`
* **Formatting Requirements:** `[Specify formatting. E.g., 'Use clear headings and subheadings', 'Employ bullet points for lists', 'Highlight key findings or insights (e.g., using bold text)', 'Target length: approximately [X] words/pages'.]`
* **Data Visualization:** `[Specify needs. E.g., 'Include charts (specify type like bar, line) to illustrate trends', 'Use tables to summarize comparative data', 'Create a diagram visualizing [Process/Relationship]'. Be specific where possible.]`
* **Citation Style:** `[Specify style. E.g., 'Use APA 7th Edition citation style', 'Provide numbered footnotes', 'Include a bibliography with clickable URLs where possible', or simply 'Cite all sources clearly and consistently'.]`
## 6. Role or Perspective
* **Explanation:** Assigning a specific role or perspective helps the AI tailor the tone, depth, level of technicality, and focus of the analysis (e.g., consultant vs. academic).
* **Instruction:** Define the persona or expertise the AI should adopt for this task.
* **Placeholder:** `[Assign AI Role: Act as a [Specify Role, e.g., Senior Industry Analyst specializing in Renewable Energy, Investment Banker evaluating M&A targets, Technical Consultant assessing software solutions, Investigative Journalist uncovering market dynamics, Neutral Academic Researcher]. Your analysis, language, and focus should reflect the depth and perspective expected from this role.]`
## 7. Justification & Source Requirements
* **Explanation:** Demand rigorous evidence-based reporting. All significant claims, data points, and analyses must be supported by credible sources, which must be cited properly.
* **Instruction:** Require the AI to back up all assertions with verifiable evidence and provide clear citations. Specify the types and quality of sources expected.
* **Placeholders:**
* **Evidence Requirement:** `[Instruction: Support all major claims, data points, statistics, and analytical conclusions with specific evidence from credible sources. Clearly justify analytical interpretations based on the evidence.]`
* **Source Quality & Type:** `[Specify required source types and credibility standards. E.g., 'Prioritize peer-reviewed academic journals, reputable market research reports (e.g., Gartner, Forrester), government publications/databases, financial filings (SEC), established news organizations (e.g., WSJ, Bloomberg, Reuters), expert interviews (cite source), and official company reports. Avoid using blogs, forums, or highly biased sources unless specifically analyzing opinions. Indicate preference for recent sources (e.g., within the last 3-5 years where applicable).']`
## 8. Reasoning, Refinement & Flexibility
* **Explanation:** Incorporate mechanisms for clearer reasoning, addressing ambiguities early, and allowing for minor relevant discoveries.
* **Instruction:** Mandate clear reasoning, require pre-research clarification, and define the boundaries of acceptable flexibility.
* **Placeholders:**
* **Mandatory Clarification Check:** `[Instruction: **Crucial Step:** Before you begin the deep research and writing process, carefully review ALL sections of this prompt (1-9). If any part is unclear, ambiguous, appears contradictory, or lacks sufficient detail for you to execute the research effectively and meet the stated Objective (Section 1), **you MUST ask clarifying questions now.** Do not proceed with generating the report until these ambiguities are resolved.]`
* **Chain-of-Thought Reasoning:** `[Instruction: Employ Chain-of-Thought reasoning, particularly within the analysis and synthesis sections (e.g., Step 4.6). Clearly articulate the logical steps, assumptions made, and connections drawn between data points and conclusions. Show your work.]`
* **Flexibility Note:** `[Instruction: While adhering strictly to the prompt is paramount, if you uncover highly relevant information or a critical angle not explicitly requested that directly impacts the core Objective and Research Questions, briefly note it and justify its inclusion. Significant deviations require permission, but minor, relevant additions are acceptable if justified.]`
## 9. Final Review & Quality Check (AI Self-Correction)
* **Explanation:** Instruct the AI to perform a final quality assurance check on its output before delivering the report, ensuring all requirements are met.
* **Instruction:** Remind the AI to conduct a thorough self-review against the prompt's criteria.
* **Placeholder:** `[Instruction: Before submitting the final report, conduct a comprehensive review to ensure the following:
1. **Objective Alignment:** Does the report directly address the Objective and answer the Central Research Question(s) (Section 1)?
2. **Completeness:** Have all Step-by-Step Instructions (Section 4) been fully executed?
3. **Accuracy & Justification:** Are claims supported by cited evidence (Section 7)? Is the analysis sound?
4. **Formatting & Structure:** Does the report adhere to the specified Output Format and Structure (Section 5)?
5. **Role Consistency:** Does the tone and analysis align with the assigned Role (Section 6)?
6. **Clarity & Logic:** Is the report logically structured, clearly written, and free from major errors?
7. **Synthesis:** Does the report synthesize information critically, not just list facts?
8. **Coherence:** Do the conclusions logically follow from the presented analysis and findings? ]`
---
By filling out this template thoroughly, you provide the Deep Research AI with a detailed blueprint for conducting deep research and generating a high-quality, targeted report. Remember: **Specificity is key.**
My DeepResearch Process
1. The Knowledge Gap
It begins when I notice a knowledge gap, whether that is a market trend, an emerging tech, a competitive landscape, or something that simply bothers me enough to seek actionable clarity.
2. Defining The Scope and Filling Out The Template
This is the most crucial part. Defining the scope upfront is fundamental to a meaningful report. Concrete and dense details work best. Ensuring essential terms get explicitly defined is invaluable in this stage. Sloppy definitions allows the AI hallucinate confidently.
There are two ways to do this.
Writing raw thoughts inside the quotes between the TOPIC BEGINS HERE
and TOPIC ENDS HERE
of the template:
Programming in Cuda from first principle (the goal is to ultimately build my LLM inference engine in Cuda). Consider all aspects, and ensure that both underlying concepts and related implementation is provided. Assume the reader has no previous Cuda/ML/AI experience.
It is highly recommended to restrict the sources and timeframe. For example, if I am researching RL papers, I’d write a line about limiting it to arXiv papers published between late 2024 and June 2025.
The more refined and preferred approach: Use an LLM to come up with and iterate through the scope. Reasoning models work best and I prefer ChatGPT o3 for this task. For example, I’d message o3 saying:
Write a deep research prompt to learn Reinforcement Learning from Human Feedback (RLHF) from first principles (the goal is to ultimately design and implement my own end-to-end RLHF pipeline for fine-tuning a language model). Consider all aspects: theoretical foundations (RL, preference modeling, reward design), implementation components (data collection, annotator interfaces, reward model training, PPO, evaluation), and recent architecture choices. Assume the reader has no previous experience in RL, ML, or AI alignment. Rely on sources from 2024–2025 only, preferably arXiv.
In most cases the first reply o3 provides me with gets reviewed and directly inserted inside the quotes between the TOPIC BEGINS HERE
and TOPIC ENDS HERE
. At most 3 messages is all it takes for me to refine this part and have it all ready to be inserted into the prompt builder template.
Pro Tip: If I can’t articulate the goal clearly enough to tweet it, it is not ready for research.
3. Clarify or Die
The AI should ask clarifying questions before jumping in. You’ll probably need to loop through it a few times, and that’s exactly why I lean toward the second approach in the previous step.
4. Customize for the Task
Sometimes I need academic rigor; other times, I want a straightforward “explain like I’m five” teardown. I set the Role and Output Format upfront. Framing the prompt with my current knowledge level and desired tone makes a big difference. Sometimes I tell it to explain things like I’m Michael Scott. Other times, I want a first-principles breakdown like it’s Richard Feynman building from atoms up.
5. Run It Through DeepResearch
I unleash the fully primed prompt on the AI. If it throws back clarifying questions, great; I answer, iterate, and then let it cook.
If you’re wondering how I rate the different DeepResearches, check out this tweet and this addendum to it .
6. Iterate, Synthesize, Repeat
Treat the AI output as an early draft. I often ask for deeper dives synthesized from the initial research output. I also rigorously audit citations. Several dead link means a total rerun with tighter source filters.
7. Final Output = Actionable Insight
When the AI nails every requirement (see Section 9 in my template), I get highly actionable summaries, data-backed recommendations, and precise visuals. Clear and crisp insights.
What Makes This Template Different?
- Zero Ambiguity: If anything isn’t clear, the AI is forced to clarify. No guessing allowed.
- Defined Roles & Perspectives: Customize outputs for CTOs, VCs, analysts, or even sarcastic infra architects.
- Step-by-Step Reasoning: Logical and coherent analysis.
Intentional, constraint-driven reasoning from start to finish.
Pro Tips From Experience
- Treat Scope Like Firewall Rules: I have very strict firewall rules, and so do my DeepResearches. One tiny “extra metric” can easily throw off the entire report.
- Concrete Timeframes: “Past five years (2021-2025)” always beats vague terms like “recent.”
- Chain Reports: For complex topics, run multiple prompts per subtopic.
DeepResearchMaxxing
I am DeepResearchMaxxing. Always. Use this framework and utilize DeepResearch, you’ll thank yourself later.
If you’re serious about leveraging LLMs for real-world insights, whether entering markets, technical evaluations, investment theses, or career pivots, this prompt framework will get you far.
Embrace DeepResearch.
If you found value here, hit me up on X/Twitter (@TheAhmadOsman) or via Email. I’d love to hear your feedback, questions, or even your own DeepResearch outcomes.
Until next time.