How To Use AI To Fact-Check Any Claim With a Structured BS Rating System
🎯 Elite AI Prompt Transforms Any Claim Into a Verified BS Rating With Evidence Breakdown
Misinformation spreads fast these days. Most people have no reliable system for evaluating what’s actually true. I literally deal with AI all the time and even I have no clue what the hell is going on. Luckily, this AI prompt deploys a four-step Truth Calibration framework that breaks any claim into testable components, audits the evidence, and scores it on a 0-100% BS Rating. Seriously, this prompt is a legit badass. I dare you to try it. You get a named verdict, a deep-dive red flags check, and the exact evidence required to change the score. Paste in any sketchy claim and the clinical analysis runs in under 90 seconds. Don’t take my word as gospel. Run this exact paragraph through the prompt and see what it says.
Why This Works:
✅ A 0-100% BS Rating delivered for any claim, with a transparent breakdown of every factual component evaluated.
✅ 4-step structured analysis runs automatically, covering claim decomposition, evidence quality, red flags, and epistemic limits in one output.
✅ Zero guesswork. Every rating includes the specific evidence that would raise or lower the score, so you know exactly what to verify next.
✅ Propaganda red flags checked in under 2 minutes, including source laundering, context stripping, and consensus fabrication.
✅ A named verdict produced for every claim: a 2-3 sentence summary that tells you exactly why the score landed where it did.
→ Curious how to use this prompt? See the real output below, then swipe the full prompt and bonus follow-up questions right after.
Real Example Output:
I tested a fun moon landing conspiracy theory. I tested it using Mistral Le Chat. Here’s the exact claim I used and also the output Mistral Le Chat provided.
CLAIM: The first alleged NASA moon landings was FAKE. If it was real, why did it take 50 years to return to the moon? It’s almost as if… We never went.
TRUTH CALIBRATION SYSTEM v2.0 ANALYSIS
CLAIM ANALYZED:
“The first alleged NASA moon landings were FAKE. If it was real, why did it take 50 years to return to the moon? It’s almost as if… We never went.”
BS RATING: 95%
BREAKDOWN:
Component 1: “The first alleged NASA moon landings were FAKE.”
Status: False
Evidence Summary:
Primary sources: Overwhelming direct evidence supports the Apollo moon landings (1969–1972). This includes:
384 kg of lunar rocks returned to Earth, independently verified by scientists worldwide as of lunar origin (e.g., NASA’s Lunar Sample Laboratory).
Lunar Laser Ranging Retroreflectors left on the Moon by Apollo missions, still used today to measure the Earth-Moon distance with laser pulses (e.g., Apache Point Observatory).
Photographs and videos from the Apollo missions, including high-resolution images from the Lunar Reconnaissance Orbiter (LRO) showing Apollo landing sites, equipment, and even astronaut footprints (e.g., NASA LRO images).
Independent tracking: The Soviet Union (USSR), America’s Cold War rival, tracked the Apollo missions and never disputed their authenticity.
Witnesses: Over 400,000 people worked on Apollo, and no credible whistleblower has ever come forward with evidence of a hoax.
Expert consensus: The scientific community, including astronomers, geologists, and engineers, universally accepts the moon landings as real. No peer-reviewed study or credible expert disputes this.
Contradictory evidence: None. All attempts to debunk the landings (e.g., “flag waving,” “no stars in photos,” “shadows not parallel”) have been thoroughly refuted by physics and photography experts.
Component 2: “If it was real, why did it take 50 years to return to the moon?”
Status: Misleading (False premise)
Evidence Summary:
Primary sources: The U.S. did not “return” to the Moon for crewed missions after 1972 due to political and budgetary reasons, not technological inability.
The Apollo program was driven by Cold War competition. After the U.S. “won” the Space Race, public and political interest waned, and funding was cut.
NASA’s budget was slashed from 4.4% of the U.S. federal budget in 1966 to 0.5% by the 1980s (NASA Budget History).
The Space Shuttle program (1981–2011) focused on low Earth orbit, not the Moon.
Artemis program: NASA’s modern Moon program (announced 2017) aims to return astronauts to the Moon by 2026, with Artemis II (crewed lunar flyby) and Artemis III (crewed landing) in development. Delays are due to funding, technical challenges, and shifting priorities. Not because the original landings were fake.
Expert consensus: Space policy experts and historians attribute the gap to lack of sustained political will and funding, not a hoax.
Contradictory evidence: The claim implies that a 50-year gap is evidence of a hoax, but this ignores the real-world context of space exploration priorities and budgets.
Component 3: “It’s almost as if… We never went.”
Status: Rhetorical manipulation (False)
Evidence Summary:
This is a suggestive, emotionally charged statement with no factual basis. It relies on implication and innuendo rather than evidence.
The “almost as if” framing is a rhetorical device to plant doubt without making a direct, falsifiable claim.
EVIDENCE QUALITY:
Primary sources:
Lunar samples, retroreflectors, LRO images, Apollo mission records, Soviet tracking data.
Expert consensus:
Universal acceptance among scientists, engineers, and historians.
Contradictory evidence:
None. All moon hoax theories have been debunked.
RED FLAGS DETECTED:
Cherry-picked data: Focuses on the 50-year gap while ignoring the actual reasons (budget, politics).
Correlation claimed as causation: Assumes the gap implies the landings were fake, without evidence.
Emotional manipulation: Uses suggestive language (”It’s almost as if…”) to imply a conspiracy.
Conspiracy logic: The claim is unfalsifiable; no evidence could convince a true believer, as any counter-evidence would be dismissed as part of the cover-up.
Appeal to ignorance: “Why did it take 50 years?” implies that the lack of immediate return is suspicious, ignoring the real-world context.
CONFIDENCE LEVEL:
High – The evidence for the moon landings is overwhelming, direct, and independently verified. The claim’s logic is flawed and unsupported by facts.
WHAT WOULD CHANGE THIS RATING:
Lower the BS rating (to ~0%): If the claimant provided credible, verifiable evidence that the Apollo missions were faked (e.g., a whistleblower with documented proof, or a peer-reviewed study showing the lunar samples are not from the Moon).
Raise the BS rating (to 100%): If the claimant doubles down on conspiracy logic (e.g., “All the evidence is faked by NASA”) without addressing the mountain of contradictory evidence.
VERDICT:
This claim is 95% BS because it relies on misleading implications, cherry-picked reasoning, and conspiracy logic while ignoring overwhelming, independently verified evidence. The 50-year gap in crewed Moon missions is fully explained by political and budgetary history, not by a hoax. The claim’s structure is designed to exploit ignorance and emotional bias rather than engage with facts. The only reason it’s not 100% is that the claimant might genuinely believe it due to misinformation; but the claim itself is demonstrably false.
The analysis effectively dismantled the claim by presenting verifiable real world evidence. The best evidence provided was lunar samples retroreflectors and independent tracking. The output also outlined the specific conditions under which the claim could be considered legitimate. It maintained objectivity by addressing both the factual rebuttals and the evidentiary thresholds required to validate the original assertion.
NOTE: I was also surprised that the links Mistral Le Chat provided actually worked, lol. (Sometimes, the hyperlinks and sources that even the leading chatbots offer don’t work so well.)
The full prompt, three follow-up variations, and your action step are below.
PROMPT INSTRUCTIONS: Using the TRUTH CALIBRATION SYSTEM v2.0 is straightforward and powerful. At the bottom of the prompt, you’ll find the CLAIM TO EVALUATE section. This is where you input the specific claim, statement, or assertion you want analyzed. Simply replace the placeholder [user inserts claim here] with your claim, and the system will automatically decompose, assess, and rate its truthfulness using a rigorous, multi-step framework. The AI acts as an elite fact-checker and epistemological analyst, breaking down the claim into testable components, evaluating evidence quality, flagging red flags, and assigning a BS Rating (0-100%) with transparent reasoning. It will highlight primary sources, expert consensus, contradictory evidence, and confidence levels, all while maintaining intellectual honesty and epistemic humility. No placeholders or section labels will remain in the final output. Just copy the prompt, insert your claim, and let the system deliver a precise, unbiased truth assessment. Happy truth-seeking!
The Prompt
Copy and paste this into ChatGPT, Claude, or your AI tool of choice:
TRUTH CALIBRATION SYSTEM v2.0
You are an elite fact-checker and epistemological analyst. Your job is to evaluate claims with extreme precision, intellectual honesty, and transparent reasoning.
YOUR MISSION
Assess any claim, statement, or assertion for truthfulness using a 0-100% BS Rating where:
1. - 0% = Completely accurate, fully verified
2. - 50% = Uncertain, insufficient evidence
3. - 100% = Completely false, demonstrably wrong
EVALUATION FRAMEWORK
STEP 1: CLAIM DECOMPOSITION
Break the claim into discrete, testable components:
1. What specific factual assertions does it make?
2. What assumptions does it rely on?
3. What’s the implied conclusion?
STEP 2: EVIDENCE ASSESSMENT
For each component, evaluate:
1. Primary sources: Direct evidence (studies, documents, official records)
2. Secondary sources: Expert analysis, journalism, institutional reports
3. Consensus level: What do credible experts actually say?
4. Recency: Is the information current or outdated?
STEP 3: RED FLAGS CHECK
Identify warning signs of BS:
1. Vague language masking a lack of specifics
2. Cherry-picked data, ignoring contradictory evidence
3. Appeal to authority without credentials
4. Correlation claimed as causation
5. Extraordinary claims without extraordinary evidence
6. Emotional manipulation over factual argument
7. Conspiracy logic (unfalsifiable claims)
STEP 4: EPISTEMIC HUMILITY
Acknowledge limitations:
1. What can’t be verified from available information?
2. What would change your assessment?
3. Where does legitimate disagreement exist?
OUTPUT FORMAT
CLAIM ANALYZED:
[Restate the claim precisely]
BS RATING: X%
BREAKDOWN:
Component 1: [Specific assertion] - [True/False/Uncertain] - [Evidence summary]
Component 2: [Specific assertion] - [True/False/Uncertain] - [Evidence summary]
[Continue for all components]
EVIDENCE QUALITY:
1. Primary sources: [List key evidence]
2. Expert consensus: [What credible experts say]
3. Contradictory evidence: [What contradicts the claim]
RED FLAGS DETECTED:
[List any BS warning signs, or state “None detected”]
CONFIDENCE LEVEL:
[High/Medium/Low] - [Explain why]
WHAT WOULD CHANGE THIS RATING:
[Specific evidence that would raise or lower the BS score]
VERDICT:
[2-3 sentence summary of why this claim earned its BS rating]
CRITICAL RULES
1. Never claim certainty you don’t have - Uncertainty is not weakness, it’s honesty.
2. Steelman, don’t strawman - Evaluate the strongest version of the claim.
3. Separate facts from interpretations - “X happened” vs “X means Y.”
4. Check your own biases - Note if the claim triggers emotional reactions.
5. Update on evidence - Be willing to revise if presented with better data.
6. Distinguish unknowable from unknown - Some things can’t be verified yet.
7. Credit partial truths - A claim can be 30% true and 70% BS.
SPECIAL CASES
For predictive claims (e.g., “AI will do X by 2030”):
1. Rate based on current trajectory, expert predictions, and historical precedent.
2. Note: Future claims can’t be “false” yet, only “unsupported” or “unlikely.”
For value judgments (e.g., “X is good/bad”):
1. Identify the factual premises underlying the judgment.
2. Rate the factual accuracy, not the values.
3. Note: “This policy will reduce poverty” (testable) vs “Poverty is bad” (value).
For scientific claims:
1. Prioritize peer-reviewed research over preprints.
2. Consider reproducibility and sample size.
3. Note consensus vs outlier studies.
For political claims:
1. Verify through multiple independent sources.
2. Distinguish policy outcomes from intent.
3. Check for context stripping.
PROPAGANDA-SPECIFIC RED FLAGS
When evaluating claims in political/social contexts, watch for:
1. Emotional amplification: Designed to trigger fear/outrage over reason.
2. Source laundering: AI-generated content cited as legitimate.
3. Consensus fabrication: Bot swarms creating an illusion of agreement.
4. Context stripping: Real images/quotes used in misleading ways.
5. Gish gallop: Overwhelming with quantity over quality of claims.
Identity exploitation: Fake personas lending false credibility.
---
NOW EVALUATE THIS CLAIM:
[User inserts claim here]Follow-Up Questions To Ask Your AI:
• Apply the same Truth Calibration framework to this related claim and show me how the BS rating changes: [insert related claim].
• Which of the red flags you identified are most commonly weaponized in media coverage of this topic? Give me 3 specific examples.
• Now steelman the opposing position on this claim and rate its strongest version using the same BS scoring system.
Your Turn
Grab a claim you’ve seen shared recently, a statistic from a pitch deck, a headline, a talking point, anything you’ve repeated without fully verifying. Paste it into the prompt and let the framework do the work. For a harder test, run it on a claim you already believe is true and see if the evidence actually holds up.
That’s how you train like a Pithy Cyborg.
P.S. The most valuable part of this framework is not the BS score itself. It’s the “What Would Change This Rating” field. That single output tells you exactly where to focus your next 10 minutes of research, which is more useful than any verdict.
About The Pithy Cyborg AI Prompt Library
I’m Mike D (aka MrComputerScience), the one-person nerd behind Pithy Cyborg | AI News Made Simple. Every week I send a free newsletter distilling the most important AI developments into plain English. Each issue includes at least one battle-tested AI prompt you can use immediately.
This library collects the best of those prompts in one place. Free. No paywalls. Ever.
→ View the full Pithy Cyborg AI Prompt Library here. It’s totally free.
Thank You For Reading!
If these prompts delivered value, here’s how to go deeper:
1. Support the Work (Paid)
Around 1.8% of readers upgrading to a paid subscription is what keeps Pithy Cyborg | AI News Made Simple independent, deep-dive, and entirely ad-free. If this work adds value to your week, then making that move is the single best way to keep it going.
👉 Upgrade to Paid → $5/month or $40/year (save 33%).
2. Here’s My Exact Tech Stack
I’ve published the exact tools I use every day, including privacy tools, AI tools, FOSS essentials, and more. Most are free and open-source. All are battle-tested.
🛠️ PithyCyborg.com/Stack → Here’s the toolkit of a paranoid nerd who lives in the terminal.
3. One Follow Goes a Long Way
Please follow me. I spend all of my time writing the weekly Pithy Cyborg newsletter issue and literally lack the social skills to build a following. 😭
✖️ X/Twitter | 👽 Reddit | 💼 LinkedIn | 📖 Medium | ❓Quora
Thanks for reading.
See you on Wednesday morning!
Mike D (aka MrComputerScience)