Detect fake news using
AI in seconds.
A production-grade machine learning system that classifies news articles as real or fake with calibrated confidence — built on TF-IDF vectorization and logistic regression.
Built with the open-source stack developers trust
Everything you need to ship fake-news detection
A complete, production-grade pipeline — from raw text input to a polished analytics dashboard.
TF-IDF + Logistic Regression
Battle-tested NLP pipeline. Stopword removal, stemming, n-gram vectorization, and a calibrated linear classifier.
Sub-200ms predictions
FastAPI backend keeps the model warm in memory. Predictions return in well under a quarter second.
Confidence intelligence
Every prediction ships with class probabilities, not just a label. Know how sure the model is.
Full analysis history
Every classification is stored to your account. Search, sort, export, or delete on demand.
Auth & rate-limited
Email + password with JWT sessions. Per-user rate limits, input validation, and secure secret handling.
Open architecture
Frontend on Vercel, ML service on Render or Docker anywhere. Swap the model without touching the UI.
Three steps from raw text to verdict
Paste an article
Drop in a headline, a paragraph, or a full article. The cleaner the input, the more accurate the prediction.
AI analyzes the text
Stopword removal, TF-IDF vectorization, and a logistic regression classifier work together to score the input.
See real-time results
Get a labeled prediction, calibrated confidence, probability bars, and automated insights — all in one card.
SaaS-grade dashboard out of the box
Track every analysis you've ever run. Watch fake-vs-real ratios shift over time. Spot patterns in confidence drift. All your detection data, beautifully visualized.
Trusted by editorial & trust teams
"We integrated this in an afternoon. The confidence scores let us route uncertain cases to human reviewers — exactly what our editorial workflow needed."
"Honestly the cleanest fake-news classifier I've shipped. The TF-IDF + LR baseline outperforms our previous BERT setup at 1/40th the cost."
"Dashboard analytics are gorgeous. My non-technical team can finally interpret model outputs without me translating."
Frequently asked questions
We use a classic NLP pipeline: text cleaning, stopword removal, TF-IDF vectorization (uni- and bi-grams), then a logistic regression classifier. It's fast, interpretable, and surprisingly competitive with much larger models on this task.