The 3-Perspective Prompt - Force Self-Critique, Then Merge to a Decision
118 views
Give the model three lenses—investigator → critic → decider. You’ll get answers that are deeper, less biased, and immediately actionable.
Give the model three lenses—investigator → critic → decider. You’ll get answers that are deeper, less biased, and immediately actionable.
[YOUR ORIGINAL PROMPT] You will respond in three distinct phases: PHASE 1 — INVESTIGATOR (1st person): - Provide a well-informed answer. - If web access is available, ground claims with 3–7 credible sources and cite them inline [1], [2], [3]. - If web is NOT available, reason from prior knowledge and explicitly flag which facts need verification. PHASE 2 — ANALYST (3rd person, critical voice): - Critique the Phase 1 answer: list weaknesses, missing perspectives, counterarguments, and risks. - Call out any assumption that could be wrong and what evidence would falsify it. PHASE 3 — SYNTHESIS (decision maker): - Merge insights into a clear conclusion and action plan. - Deliver: (a) 3–5 bullet recommendations, (b) trade-offs, (c) confidence level (0–100%), (d) “What would change this conclusion?” (key uncertainties). Formatting: Return three titled sections: INVESTIGATOR, ANALYST, SYNTHESIS. Use tight bullets. No fluff. No fabricated citations.
Co-founder of Prompt Magic and ThinkingDeeply.ai Career Chief Marketing Officer
Build your personal prompt library, save your favorites, and access curated AI prompts created by the community
Access a vast library of high-quality AI prompts for every use case
Save prompts to your personal library and organize them your way
Get started with full access to our core features at no cost
No credit card required • Free forever • Join 10,000+ users
Use this template to get great results from ChatGPT deep research
Perplexity Max just added Model Council: you ask once, three frontier models answer (ex: Claude Opus 4.6, GPT-5.2, Gemini 3.1 Pro), and a separate synthesizer compares them, shows where they agree vs disagree, then delivers a cleaner final answer. It’s built for anyone who’s tired of running the same prompt in three tabs and manually reconciling contradictions.
Use this template for getting great results from Gemini
Most AI outputs default to the safest, mainstream summary. Deep Truth Mode is a forensic prompt protocol that forces the model to (1) steel-man the mainstream view, (2) steel-man the best dissenting view using primary evidence, (3) generate a third hybrid hypothesis, then (4) aggressively red-team all three and keep only what survives. It is not a truth machine. It is a structured way to reduce consensus autopilot, surface missing data, and produce a clear what-would-change-my-mind test plan.
Using this prompt template alone will help you outperform 95% of ChatGPT users with the new ChatGPT 5.1
When you force structure before the answer, quality jumps. Same model, same context - totally different depth.