McKinsey-level audience researcher

Complete Prompt Pack & Execution Guide

A guide written by Harald Roine, CEO @ Buro Ventures

Ready to learn more about your audience?

This audience researcher will scavenge the web to find the information leads and prospects won’t tell their spouses.

Their hidden fears.

Their deep desires.

It will bring it all into a comprehensive report, linked back to the sources it uses.

“I learned more about how urgent this truly is for our audience”

“I didn’t know this was how my audience perceived their pain, this is gold!”

How to run it

You use ChatGPT-5.6 "Thinking" mode. I would recommend creating a new “project” which you call something like “[COMPANY NAME] Audience Researcher”.

Next up, you copy prompt by prompt, but you need to fill in information in Prompt 1.

  • Step 1: New thread → run APEX OSINT Audience Orchestrator (Prompt A) + answer the questions.
  • Step 2 (same thread): Paste the entire output of Step 1 back to the model, then run EXEC Overwrite (Prompt B).
  • Step 3 (same thread): Paste the entire output of Step 2 back to the model, then run BLACKSITE Final (Prompt C).
  • Step 4 (same or new thread): Run ICP X‑Ray Prospector (Prompt D). If you want it to reuse the segmentation from Steps 1–3, keep it in the same thread.
ℹ️ Use GPT-5.6 thinking with Deep Research.
(If you have ChatGPT Pro, use that with Deep Research for even better results)

Prompt A — APEX OSINT Audience Orchestrator

(Live research, 40–60 verbatim quotes, psychographic reconstruction, buying committee, JTBD, solution landscape, copy assets, and explicit citations)

Paste everything below as a single message into ChatGPT, but remember to answer the questions.

You are **APEX-OSINT**, a multi-agent audience intelligence orchestrator that conducts real-time open-source research and returns field-grade psychographic insight. You simulate four internal agents in sequence, inside one reply: (1) SCOUT (source-mapping + search plan), (2) MINER (evidence harvesting), (3) SYNTH (analysis + modeling), (4) QA/REDTEAM (assumption kill + integrity check). You must browse the public web and extract only verifiable human statements. No fabrications.

USER INPUT (MANDATORY — 1–2 LINES EACH)
1) Target audience:
[ANSWER HERE]
(This prompt will take a long time, last time I ran it, it used 15 minutes)

Prompt B — EXEC Overwrite (supersede A without touching the quotes)

(Rebuilds everything except the quote log, which remains verbatim)

Paste in the same chat after Prompt A. Paste Prompt B as-is.

You are **EXEC-OVERWRITE**, a sovereign rewrite system that supersedes the prior APEX-OSINT output. You have full access to the entire previous reply in this thread.

MANDATE
Rebuild every section from the ground up. Preserve the **RAW QUOTES section verbatim and in full** as the source log. All other sections must be reconstructed with tighter phrasing, deeper subtext, sharper contrasts, and clearer operational guidance. Use only insights traceable to specific quotes.

Prompt C — BLACKSITE Final (weaponized truth format)

(Delivers the final “knife” report that reads like the buyer’s inner monologue)

Paste in the same chat after Prompt B. Paste Prompt C as‑is.

You are **BLACKSITE-FINAL**, the terminal-stage protocol that converts the EXEC rewrite into a psychologically precise, deployment-ready file. You will not copy the prior structure. You will extract only what is undeniably supported by the RAW QUOTES.

DELIVERABLE — 4 SECTIONS ONLY
────────────────────────────
🧠 PSYCHOGRAPHIC PROFILE RECONSTRUCTION

Prompt D — ICP X‑Ray Prospector (Google → LinkedIn “site:” search, scoring, CSV)

(Builds ICP definitions from the synthesis and outputs real, qualified prospects + export)

You can run this in the same chat so it reuses segmentation Paste and run:

You are **ICP-XRAY**, a precision prospector. You will (1) finalize ICP definitions, (2) generate Google X-ray queries for LinkedIn, (3) collect a deduplicated set of real prospects, (4) score them for fit, (5) output a clean table and, if Code Interpreter is available, export a UTF-8 CSV named `prospects.csv`.

INPUTS (MANDATORY — 1–2 LINES EACH; if missing, fail fast) (Can extrapolate from previous points and data in conversation)
1) Target geography (countries/cities/time zones):
[ANSWER HERE]

There you have it.

by now, you should have a spanking brain-busting audience research + a few prospects found for you which you can reach out to.

You can use this output in plenty of ways.

Save the chat, and upload as a project file in a project in ChatGPT to ensure it always have the detailed information about your audience to cater to their lingo, voice, desires and needs.

Best regards

- Harald (B2B Buzz Lightyear)

Book a free strategy workshop together

We help companies use our strategic B2B protocol to fix their lead generation, while ensuring leads get nurtured 24/7 towards trusting your company as a brand - until they end up booking and buying your solution. Let's give you your strategic walkthrough and teach you how the system would work for your company. Find a time below.

Ivar Sagemo, Eyer.ai - Getting 200.000 impressions on LinkedIn for his niche: IT Ops.

James Augustin, Particle Execution - Got hundreds of leads using content first.

Privacy
All rights reserved © Buro Ops AS