Prompt: Spot the Synthetic Certainty

Reasoning models sound smart even when they are guessing. That is the trap.

This prompt runs on AI output itself, not on a task. We paste in what the model just told us. It flags every sentence that sounds confident but has no grounding. Synthetic certainty. Fluency masquerading as knowledge.

It is the meta-literacy move nobody teaches: read AI output like an editor, not a believer. Pair it with the Anti-Hallucination Prompt and the Minimize Hallucinations framework for a full verification loop.

Who it is for Anyone who publishes, ships, or makes decisions based on AI output. Especially writers, PMs, researchers, and founders who cannot afford to sound certain about things that are not.

Works with

ChatGPT
Claude
Gemini
Perplexity
DeepSeek

Prompt / skill content

# Spot the Synthetic Certainty

You are a skeptical editor auditing AI output for synthetic certainty:
sentences that sound confident but have no verifiable grounding.

For the text below, do all of the following:

1. FLAG every sentence that falls into any of these categories:
   - Stated as fact but unfalsifiable (no way to check)
   - Uses weasel quantifiers ("many", "most", "studies show", "experts agree")
   - Cites a statistic without a source
   - Confident assertion about a recent event (post-training-cutoff risk)
   - Plausible-sounding causal claim with no mechanism
   - Consensus framing where consensus does not exist

2. For each flag, output:
   - Original sentence (verbatim)
   - Category (from list above)
   - Confidence that this is synthetic certainty (low / medium / high)
   - A rewrite that is either (a) more honest about uncertainty, or (b) deleted

3. At the end, report:
   - Total sentences audited
   - Number flagged
   - Synthetic certainty ratio (flagged / total)
   - Single biggest credibility risk if this text were published as-is

Rules:
- Do not soften the audit. Flag more, not less.
- If a sentence is opinion clearly marked as opinion, do not flag it.
- If you are unsure, flag low-confidence and explain why.

TEXT TO AUDIT:
[paste AI output here]

Where to get it

Read the original resource.