AI Provenance Tracker
Upload, paste, or analyze URL and generate an explainable report fast. Built for fact-checkers and newsroom verification workflows.
Tag @whoisfake or DM a public Instagram link, then fall back to URL detection when a platform post does not expose public media.
GPT-4, Claude, Llama detection via perplexity, burstiness, and DistilRoBERTa
DALL-E, Midjourney, Stable Diffusion detection via FFT and CNN analysis
AI speech detection via spectral flatness, dynamic range, and zero-crossing
Deepfake detection via container signatures and byte-pattern analysis
Try It
Upload/Paste/URL -> Explainable report.
Methodology limits: results are probabilistic and should be treated as evidence support, not a final verdict. Experimental modalities: social URL resolution depends on public OG metadata and may not work for private/auth-required pages. Instagram public auto-replies are limited to our own media comments; third-party mentions fall back to DM or manual URL analysis.
How It Works
A multi-signal detection engine that analyses content through multiple independent techniques and combines them into an explainable confidence score.
Multi-Signal Analysis
Combines perplexity, burstiness, vocabulary distribution, and ML classifiers into a weighted confidence score — not just a binary yes/no.
Explainable Results
Every detection returns a signal-by-signal breakdown showing exactly why content was flagged, making results transparent and auditable.
Multi-Provider Consensus
Aggregates results from internal detectors and external providers (Copyleaks, Reality Defender, C2PA) with configurable weighting.
Public Benchmark
Reproducible evaluation framework with ROC-AUC, calibration ECE, and Brier score across detection, attribution, and tamper robustness tasks.
Deploy Anywhere
Docker Compose for local, Helm charts for Kubernetes, Terraform IaC for AWS — production-ready with API/worker process splitting.
Open Source
MIT licensed. Full detection methodology documented. Self-host, audit, and extend with complete transparency.
Detection Pipeline
Run Verification Workflows
Start detection first, then review dashboard/history if needed.