Fact-Checking and Hallucination Prevention
Grounding-based verification: every claim must trace to a retrieved source.
Fact-Checking and Hallucination Prevention is a free AI Agents lesson on CoddyKit — lesson 4 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the AI Agents learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.
The Hallucination Problem in Research Agents
LLMs can generate plausible-sounding claims with no basis in the retrieved sources. In a research agent, this is especially dangerous because the output looks authoritative and is presented with citations that may not actually support the claim.
Hallucination prevention must be a first-class concern.
Grounding: Every Claim Traces to a Source
The core principle of grounding: every factual claim in the final output must be traceable to at least one retrieved document. Claims that cannot be traced are either hallucinated or unsupported — both are unacceptable in a research report.
def check_grounding(claim: str, retrieved_docs: list[dict]) -> dict:
doc_texts = '\n\n'.join(
f'[DOC {i+1}] ({d["url"]})\n{d["text"][:800]}'
for i, d in enumerate(retrieved_docs[:5])
)
return {
'claim': claim,
'docs': doc_texts,
'grounded': None # to be filled by LLM verifier
}
if __name__ == '__main__':
docs = [{'url': 'https://example.com/geo', 'text': 'Paris is the capital of France.'}]
result = check_grounding('Paris is the capital of France.', docs)
print('Claim:', result['claim'])
print('Supporting docs used:')
print(result['docs'])
LLM-as-Verifier Pattern
Use a separate LLM call — the 'verifier' — to judge whether a claim is supported by the provided sources. This creates a check-and-balance where the generator and verifier are independent calls.
import openai, json
client = openai.OpenAI(api_key='YOUR_OPENAI_KEY')
def verify_claim(claim: str, source_texts: list[str]) -> dict:
sources_block = '\n---\n'.join(source_texts[:3])
prompt = (
f'Is the following claim directly supported by the provided sources?\n'
f'Claim: "{claim}"\n\n'
f'Sources:\n{sources_block}\n\n'
f'Return JSON: {{\n'
f' "supported": true/false,\n'
f' "confidence": 0.0-1.0,\n'
f' "reason": "one sentence explanation"\n'
f'}}'
)
resp = client.chat.completions.create(
model='gpt-4o',
messages=[{'role': 'user', 'content': prompt}],
response_format={'type': 'json_object'}
)
return json.loads(resp.choices[0].message.content)Confidence-Gated Output
Set thresholds on the verifier's confidence score. Claims above 0.85 are published. Between 0.5-0.85, they are published with a disclaimer. Below 0.5, they are excluded.
INCLUDE_THRESHOLD = 0.85
DISCLAIMER_THRESHOLD = 0.50
def gate_claim(claim: str, source_texts: list[str]) -> dict:
result = verify_claim(claim, source_texts)
conf = result.get('confidence', 0.0)
supported = result.get('supported', False)
if not supported or conf < DISCLAIMER_THRESHOLD:
return {'action': 'exclude', 'claim': claim, 'reason': result.get('reason')}
elif conf < INCLUDE_THRESHOLD:
return {
'action': 'include_with_disclaimer',
'claim': f'[LOW CONFIDENCE] {claim}',
'reason': result.get('reason')
}
else:
return {'action': 'include', 'claim': claim}Catching Specific Hallucination Patterns
Certain claim patterns are high-risk for hallucination: exact statistics (percentages, dollar amounts), specific dates, named individuals, and causal relationships. Apply extra scrutiny to these.
import re
def is_high_risk_claim(claim: str) -> bool:
patterns = [
r'\d+\.?\d*\s*%', # percentages: 9.1%
r'\$\s*\d+', # dollar amounts
r'\b(January|February|March|April|May|June|July|August|September|October|November|December)\s+\d{4}', # dates
r'\b[A-Z][a-z]+\s+[A-Z][a-z]+\s+(said|stated|argued|claimed)', # named quotes
r'(caused?|led to|resulted in)', # causal claims
]
return any(re.search(p, claim) for p in patterns)
print(is_high_risk_claim('Inflation hit 9.1% in June 2022')) # True
print(is_high_risk_claim('Inflation was elevated')) # FalseThe Decomposed Verification Loop
For long sections, decompose the text into individual claims, verify each claim independently, then reconstruct the section with only verified claims.
def decompose_into_claims(section_text: str) -> list[str]:
prompt = (
f'Break this text into individual verifiable factual claims.\n'
f'Each claim should be a single sentence containing exactly one fact.\n'
f'Return JSON: {{"claims": ["..."]}}\n\n'
f'TEXT:\n{section_text}'
)
resp = client.chat.completions.create(
model='gpt-4o-mini',
messages=[{'role': 'user', 'content': prompt}],
response_format={'type': 'json_object'}
)
return json.loads(resp.choices[0].message.content).get('claims', [])Reconstructing Verified Text
After gating each claim, reconstruct a clean, coherent section from only the verified claims. Exclude low-confidence ones and rewrite for readability.
def reconstruct_section(verified_claims: list[str],
section_name: str) -> str:
claims_text = '\n'.join(f'- {c}' for c in verified_claims)
prompt = (
f'Rewrite these verified facts as a coherent {section_name} section.\n'
f'Do NOT add any new information not present in the claims.\n'
f'Only use the facts provided.\n\n'
f'VERIFIED CLAIMS:\n{claims_text}'
)
resp = client.chat.completions.create(
model='gpt-4o',
messages=[{'role': 'user', 'content': prompt}]
)
return resp.choices[0].message.contentDetecting Numeric Drift
LLMs sometimes subtly alter numbers from the source (e.g., 9.1% becomes 9.2%). Extract numbers from both the claim and the source text and compare them explicitly.
import re
def extract_numbers(text: str) -> list[float]:
matches = re.findall(r'[-+]?\d*\.?\d+', text)
return [float(m) for m in matches]
def check_numeric_drift(claim: str, source_text: str,
tolerance: float = 0.01) -> bool:
claim_nums = extract_numbers(claim)
source_nums = extract_numbers(source_text)
for cn in claim_nums:
found_match = any(abs(cn - sn) <= tolerance for sn in source_nums)
if not found_match:
return True # numeric drift detected
return False
print(check_numeric_drift(
'Inflation reached 9.2% in June 2022',
'The CPI rose 9.1 percent in June 2022'
)) # True — 9.2 vs 9.1Attribution Audit
After generating a full report, run an attribution audit: for each sentence in the report, confirm it can be attributed to one of the sources. Flag any sentence that has no matching source snippet.
def attribution_audit(report_text: str, sources: list[dict]) -> list[str]:
sentences = [s.strip() for s in report_text.split('.') if len(s.strip()) > 20]
unattributed = []
for sentence in sentences:
found = False
for src in sources:
if any(word in src.get('text', '') for word in sentence.split()[:5]):
found = True
break
if not found:
unattributed.append(sentence)
return unattributed
# Flag unattributed sentences for manual review or re-verification
if __name__ == '__main__':
report = ("Water boils at 100 degrees Celsius at sea level. "
"The moon is made primarily of green cheese according to this document.")
sources = [{'text': 'Water boils at 100C (212F) at standard atmospheric pressure.'}]
unattributed = attribution_audit(report, sources)
print('Unattributed sentences (need manual review):')
for s in unattributed:
print(' -', s)
Hedging Uncertain Claims
Not everything can be verified to high confidence. Instead of excluding borderline claims entirely, use hedging language: 'Some analysts suggest...', 'According to X...', 'It has been reported that...'. This preserves information while signaling uncertainty.
def hedge_claim(claim: str, confidence: float) -> str:
if confidence >= 0.85:
return claim # state as fact
elif confidence >= 0.65:
hedges = ['Some sources suggest', 'According to available evidence',
'Analysts have noted']
return f'{hedges[hash(claim) % len(hedges)]}, {claim.lower()}'
else:
return f'It has been reported (with low confidence) that {claim.lower()}'
print(hedge_claim('Inflation peaked at 9.1%', 0.90))
print(hedge_claim('Supply chain disruptions contributed to inflation', 0.72))
print(hedge_claim('Specific policy caused inflation', 0.45))Hallucination Rate Monitoring
Track hallucination rate over time: what percentage of generated claims fail verification? Set an alert threshold. If it spikes, your prompt engineering or retrieval quality has degraded.
verification_log = [] # In production: a database
def log_verification(claim: str, supported: bool, confidence: float):
verification_log.append({
'claim': claim[:100],
'supported': supported,
'confidence': confidence
})
def hallucination_rate() -> float:
if not verification_log:
return 0.0
failed = sum(1 for v in verification_log if not v['supported'])
return failed / len(verification_log)
def check_hallucination_alert(threshold: float = 0.15):
rate = hallucination_rate()
if rate > threshold:
print(f'ALERT: Hallucination rate {rate:.1%} exceeds threshold {threshold:.1%}')
return rate
if __name__ == '__main__':
log_verification('The sky is blue', True, 0.95)
log_verification('The moon is made of cheese', False, 0.2)
log_verification('Water boils at 100C at sea level', True, 0.99)
rate = check_hallucination_alert(threshold=0.15)
print(f'Hallucination rate so far: {rate:.1%}')
What is the primary purpose of the 'LLM-as-verifier' pattern?
The LLM-as-verifier is a key architectural pattern in hallucination prevention. Understanding what it checks and why it is done as a separate call is essential.
Hallucination Prevention Recap
Prevent hallucinations through: grounding checks (every claim traces to a source), LLM-as-verifier (separate call judges support), confidence gating (exclude below-threshold claims), numeric drift detection, hedging language for borderline cases, and hallucination rate monitoring to detect regressions.
Frequently asked questions
Is the “Fact-Checking and Hallucination Prevention” lesson free?
Yes — the full text of “Fact-Checking and Hallucination Prevention” is free to read here on the web, and the AI Agents course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the AI Agents course, upgrade to CoddyKit PRO.
What will I learn in “Fact-Checking and Hallucination Prevention”?
Grounding-based verification: every claim must trace to a retrieved source. You practise AI Agents with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.
Do I need any experience to start AI Agents?
No prior experience is required. AI Agents on CoddyKit is structured for beginners through advanced learners; this is — lesson 4 of 4, so you can start here or from the beginning and move at your own pace.
How long does the “Fact-Checking and Hallucination Prevention” lesson take?
Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.
Can I write and run code in this AI Agents lesson?
Yes. Every AI Agents lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.
All lessons in this course
- Multi-Step Research Loop Design
- Source Verification and Citation
- Structured Report Generation
- Fact-Checking and Hallucination Prevention