AI Research System: Source-Graded Research Prompts with Confidence Labels

What is inside

• 35 prompts across 9 phases
• Claim definition and competitor registry method
• Four confidence labels that are never upgraded
• Contradiction log, audit-finding triage and a refresh cadence
• PDF guide with a Start Here page and three quick-win prompts
• PROMPTS.md and MASTER-INSTRUCTIONS.md you can paste straight into Claude, ChatGPT or Gemini
• Every prompt laid out the same way: purpose, when to use, inputs, prompt text, sample output, verify step, failure and fix, variants
• “The 50 Practices Behind This Pack” appendix and a 25-check quality checklist

Who it is for

Consultants, writers and analysts who need research that survives a challenge.

Why it is different

A figure with no opened source is cut, rephrased or disclosed. That one rule changes the quality of everything that follows.

Important notice. AI prompts produce drafts, so review every output before you publish, send, submit or pay. The prompts were written and reviewed in October 2026 and were not formally tested across assistants. Statements about outside platforms, standards and organizations were checked on 2026-10-25 against the sources named in the pack and anything that could not be confirmed was removed, but rules change, so confirm current terms before you act. No results, income, rankings, approvals or sales are promised. This is educational material, not legal, tax or financial advice. Never paste passwords, keys, card numbers or confidential data into an assistant. Not affiliated with any platform named. Created with AI assistance.