Stop misinformation before it spreads.
Get the truth instantly.
FactCheckAI analyzes claims in real-time using advanced AI models and cross-references millions of trusted sources. Know what's true while you browse.
"5G towers spread coronavirus through high-frequency radio waves."
Multiple international public health authorities and telecommunications engineering bodies confirm non-ionizing electromagnetic radiation cannot synthesize or transmit biological viral matter.
Check a claim in seconds.
Paste a statement or explore structured evaluations generated through hybrid retrieval and citation cross-validation.
No official WHO decree matches this assertion, but ongoing regulatory evaluations regarding excessive synthetic megadoses create ambiguous colloquial reporting across news wires.
"Tolerable Upper Intake Levels for Vitamin D reaffirmed; no systemic bloodstream toxicity detected under standard therapeutic guidelines."
"Hypercalcemia risks are limited to hyper-megadose cases (>50,000 IU/day over prolonged intervals). No broad categorization of toxicity exists."
How FactCheckAI verifies information
Every evaluation traverses a 4-step forensic verification loop designed to prevent hallucinations and establish an immutable chain of custody for sources.
Claim Extraction
The user selects text from any live webpage or submits a raw string. The local client isolates entity boundaries and context sentences.
ML Analysis
Fine-tuned RoBERTa models evaluate linguistic markers, sensationalist bias, emotional appeal, and decompose compound claims into atomic assertions.
Evidence + RAG
Reciprocal Rank Fusion simultaneously queries BM25 indices and dense pgvector embeddings over millions of peer-reviewed articles and accredited news wires.
Final Assessment
Evidence signals are synthesized with chain-of-thought verification. Uncertainty is calibrated and output alongside verifiable source citations.
Evidence from trusted institutions
Our hybrid retrieval engine queries over 7 million documents, ensuring every verdict is grounded in verifiable, authoritative sources.
Peer-Reviewed Studies
Cross-referenced from PubMed, arXiv, IEEE Xplore, and institutional repositories.
Accredited News Wires
Vetted journalism from Reuters, AP, BBC, and fact-check consortiums like Poynter IFCN.
Government & NGO Data
Primary data from WHO, CDC, NOAA, EPA, and UN specialized agencies.
Built on open-source AI research
Transparent, reproducible, and peer-reviewed architectures power every component—no black-box magic.
Fine-tuned transformer models (RoBERTa, BERT) specialized for stance detection, claim decomposition, and bias identification.
Hybrid BM25 + dense vector search with reciprocal rank fusion ensures both keyword precision and semantic recall.
Probabilistic calibration using temperature scaling and ensemble disagreement metrics to surface epistemic uncertainty.
Check claims without leaving your tab
Select any text on any webpage. Right-click to invoke FactCheckAI. Get instant verdicts with sourced evidence—no tab switching required.
Breaking: New Study Claims...
According to researchers at the university, "synthetic vitamin supplements are now considered toxic by WHO", sparking controversy in the medical community.
No WHO declaration matches this claim. Cross-referenced with 3 official sources showing no toxicity classification exists.
Instant Context Menu
Right-click any selected text to trigger instant fact-checking without opening new tabs.
Visual Highlights
Suspicious claims are automatically highlighted with color-coded severity indicators.
Session History
Review all fact-checks from your browsing session with full evidence trails.
Honest about what we don't know
Unlike binary fact-checkers, FactCheckAI surfaces uncertainty when evidence is conflicting or insufficient. Confidence scores are calibrated to reflect epistemic limitations.
REAL
Strong consensus across multiple high-authority sources with no conflicting evidence.
UNCERTAIN
Conflicting evidence, ambiguous framing, or insufficient authoritative coverage.
FAKE
Direct contradiction by multiple credible sources with strong empirical refutation.
Calibrated Confidence
Our models undergo post-hoc calibration to ensure confidence scores accurately reflect true prediction accuracy. When we say 90% confidence, we're right ~90% of the time—not inflating certainty to seem more authoritative.
Epistemic Humility
Emerging claims, niche scientific debates, or politically contested topics often lack clear consensus. We flag these as UNCERTAIN rather than forcing a binary verdict—honoring the complexity of real-world information.
Every verdict comes with receipts
Full audit trails show exactly which sources were consulted, how confidence was calculated, and which reasoning chain produced the final verdict.
Claim Ingestion
t=0msUser-selected text captured with full context
ML Classification
t=42msRoBERTa stance model: 0.87 fake probability
Evidence Retrieval
t=128ms23 sources retrieved via hybrid BM25+vector
Citation Validation
t=165ms3 sources verified, 2 contradictory dismissed
Uncertainty Calibration
t=182msFinal confidence: 87.4% (calibrated)
Full Source Citations
Every verdict includes clickable citations to original source documents with relevance scores.
Reasoning Chain Export
Download complete JSON logs of model inputs, retrieval results, and decision logic.
User Feedback Loop
Report incorrect verdicts to improve models. All feedback is human-reviewed and transparent.
Built with transparency, powered by community
Every model, algorithm, and data pipeline is openly documented. Community contributions drive continuous improvement.
Join the community
Contribute to models, suggest features, or help improve documentation. All skill levels welcome.
Questions & Answers
Common queries about how FactCheckAI works, privacy, accuracy, and contribution guidelines.
Unlike binary fact-checkers, we surface uncertainty when evidence is ambiguous. We use hybrid ML+retrieval instead of purely manual fact-checking, and we provide full transparency with source citations and confidence calibration.
Ready to stop misinformation?
Get the extension now and start verifying claims instantly. It's free, works on all major browsers, and respects your privacy.