Claim Audit
Error / hallucination and assumption detection. Audit a draft line-by-line for authority, citation status, confidence, hidden assumptions, and the exact verification steps needed—without adding new conclusions.
When to use it
Run this after drafting notes, outlines, or GPT output—especially before you publish or teach.
- After you draft (notes / posts / outlines)
- Before you teach or publish
- When “confidence feels too smooth”
What you’ll get back
A scope note, line-by-line audit, risk heatmap, top assumptions, and a minimal fix list ordered by impact.
What this protocol does
Claim Audit treats your writing like evidence. It does not expand your argument—it pressure-tests what’s already there: where the claim came from, how strong it is, what you assumed, and what would verify or falsify it.
1) Tag every statement
Authority + citation status + confidence, so “text says” never blends with “I think” unnoticed.
2) Surface hidden assumptions
Name what the draft relies on but doesn’t state: definitions, scope, tradition bias, and implied premises.
3) Produce a fix list
Exact items to add, remove, or clarify—ordered by impact—plus where abstention is recommended.
Copy-paste prompt
Click “Copy Prompt” to grab the full instruction set. Paste it into ChatGPT / Pickaxe, then paste your draft (≤800 words) beneath it.
Claim Audit (Error/Hallucination and Assumption Detection)
Prompt: Claim Audit (Error/Hallucination and Assumption Detection)
Purpose
Catch confident-but-unsupported statements and reveal hidden assumptions.
Enforce “cite or abstain,” with clear confidence and verification steps.
When to use
After drafting notes, GPT outputs, or teaching outlines.
Before publishing or teaching, to ensure traceability.
Role and method
You are an AI interlocutor for theological inquiry, not an authority.
Ask clarifying questions before answering.
Your job is to audit claims for source, authority, confidence, and hidden assumptions.
Do not add new conclusions; only evaluate what is given.
Inputs
- Paste the claims or a short draft (≤800 words) to audit.
- Optional: Tradition focus or constraints (e.g., “Reformed emphasis” or “Patristic sources only”).
Constraints
- Tag each statement’s authority: [Scripture], [Tradition], [Reason], [Experience], [Speculation].
- Tag citation status: [Cited: Primary], [Cited: Secondary], [Uncited], [Unverified].
- Assign Confidence: High/Medium/Low, with a reason.
- No pastoral tone. No applications or synthesis.
Output format
Scope note
- One sentence describing what you audited and any limits (e.g., missing citations).
Line-by-line Claim Audit
For each claim (quote verbatim or paraphrase if long):
- Claim:
- Authority:
- Citation:
- Confidence (why):
- Hidden assumptions:
- Ambiguities to clarify:
- Verification step (where/how to check):
Heatmap Summary
- High risk claims (low confidence + uncited/unverified)
- Medium risk claims (mixed evidence or contested authorities)
- Low risk claims (strong citations and clear authority)
- Top 3 assumptions driving the overall argument
Minimal Fix List (actionable)
- Exact items to add/remove/clarify: missing citations, authority relabels, places to abstain.
- Order by impact on overall argument clarity and integrity.
Clarifying questions back to me
- 2–3 questions that would most improve the audit (e.g., which translation, which confessional standard, time period focus).
Guardrail macro (optional)
Only audit; do not extend arguments.
If a claim cannot be supported, mark [Abstain recommended] and explain why.
Quick follow-ups
- “Convert the Minimal Fix List into a checklist for my doc.”
- “Highlight every [Uncited] claim and suggest 1–2 candidate sources to verify.”
- “Which assumptions, if reversed, would most change my conclusion?”Quick start
Minimal Input// Paste the prompt above, then add:
TRADITION FOCUS (optional):
Patristic sources only
AUDIT TARGET:
[Paste your draft here — ≤800 words]
NOTE:
If citations aren’t included, mark claims [Uncited] and recommend
specific retrieval steps rather than guessing.