A structured prompt for fact-checking and detecting AI-generated posts

Found a well-structured prompt template for auditing social media posts — especially geopolitical content that may be AI-generated. It runs two separate analyses:

Analysis 1: Line-Item Fact Check Extract every specific factual claim (military units, ship names, troop numbers, casualty figures, oil prices, named officials, institutional attributions) and verify each against at least two independent sources. Verdicts:

• CONFIRMED — verified by multiple sources
• DISTORTED — approximately correct but misstated in degree/detail
• UNVERIFIED — no source found for the specific claim as written
• FRAMING — interpretation presented as declarative fact (e.g. “The call was not about peace. It was about oil.”)

Analysis 2: AI Generation Probability Score 6 indicators from 0–10:

1. Structural uniformity — follows high-volume AI content templates (dramatic opener → hardware inventory → price data → strategic interpretation → literary closer)
2. Sourcing absence — interpretive claims without named analysts
3. Cinematic detail — narrative flourishes connecting unrelated facts through implied causation
4. Hedge avoidance — unqualified declarative language without uncertainty markers
5. Volume consistency — posting frequency vs. realistic human research timelines
6. Comment engagement — does the author engage with factual challenges, or post and move on?

Takeaway: The “FRAMING” verdict category is particularly useful. A lot of viral geopolitical content mixes real facts with interpretive leaps stated as though they’re facts too. Splitting these apart is the key skill — both for humans reading the post and for AI auditing it.

Credit: @TheBlackWallaby