free-prompts — 5 AI OSINT prompts that make the model collect real data instead of hallucinating it

Ask ChatGPT or Claude to investigate a target and it does one of two things: it either refuses on ethics grounds, or it produces confident-sounding output that has nothing to do with real public data. Both outcomes are useless for actual OSINT work.

The issue is not the model — it’s the prompt. Without structure, the AI improvises. With structure, it executes.

These 5 prompts impose that structure, one per stage of an investigation: scope → collect → pivot → verify → document. Each prompt tells the model exactly what to return, in what format, and what not to do. The result is reproducible, auditable output you can act on.

• Junior OSINT analysts who want a repeatable investigation method
• Pentesters and bug bounty hunters doing recon with AI tooling
• Investigators running AI-assisted due diligence on people, companies, or IPs
• Anyone who has tried using ChatGPT or Claude for recon and got garbage output

Works with Claude, ChatGPT, or any AI assistant. Supercharged when paired with OpenOSINT — open-source OSINT framework with 14 tools and an MCP server.

• 5 tested prompts, one per investigation stage — each designed to produce structured, non-hallucinated output
• The method behind them — a short explanation of why each is structured the way it is, so you can adapt it
• Ethics note built in — every prompt includes an authorisation check so the model flags concerns before proceeding
• Instant PDF — enter $0 on Gumroad, no card required, download immediately

Prompt 1 verbatim — the scope stage. Every investigation starts here before any tool runs:

You are a structured OSINT analyst working on authorised investigations only.
Before any tool runs, produce a scope document for the target below.

Target: {{TARGET}}

Return exactly this structure — no prose, no filler:

## Target profile
- Type: [email / username / IP / domain / phone / full name]
- Input as given: {{TARGET}}
- Investigation goal: [one sentence — what question are we actually answering?]

## Tool plan
List each tool you intend to call, in order, with one-line justification:
1. <tool_name> — reason
2. <tool_name> — reason
(add as many as needed)

## Out-of-scope
What this investigation will NOT cover, and why.

## Ethics check
Flag any legal or ethical concern before proceeding.
If none: "No concerns identified."

Do not call any tools yet. Return only the scope document above.

The remaining 4 prompts cover: structured collect, cross-source pivoting, multi-source verification, and final report assembly. All in the PDF.

Preview: one line from Prompt 4 (pivot stage — in the full pack only):

[Paid — Prompt #4 of 30+: Username Pivot]
Given a confirmed username from step 2, expand to connected accounts:
██████████████████████████████████████████████
Cross-reference ███████████████ to surface █████████████████████.
Then check ████████████ for ███████████████████████████████.
Return findings as ████████████████████████ with confidence scores.

Enter $0 on Gumroad. No card, no account required. The PDF downloads immediately.

From the maker of OpenOSINT, an open-source OSINT framework used by security researchers and analysts worldwide. No spam — unsubscribe anytime.

The free set walks the investigation loop once. The AI OSINT Prompt Pack has 30+ prompts covering every investigation type: email, username, domain, IP, phone, company due diligence, image analysis, verification patterns, and final reporting — plus the complete methodology and an ethics and legal primer.

Buy the full pack ($29) ↗