i
Research-grade tool — not a fact-checker. This is a statistical detector built for media-literacy research. It does not verify claims against reality, is not a moderation or censorship aid, and its verdicts are never definitive. Always read predictions alongside their confidence.
Stop Propaganda
Check news in 1 click

Check the news for propaganda.

Paste a tweet, a headline, or any paragraph you find suspicious. In about half a second you’ll know whether it reads as propaganda — and how confident the model is. As a bonus, it names the exact technique and shows where it appears.

Text to analyze
Model
0

Glossary

The vocabulary this tool teaches. Five high-level intents, seventeen fine-grained techniques, and twenty-two entity-framing roles — after the Solopova schema (NLP4PI 2025).

Five intents

Seventeen techniques

Entity-framing roles

About this tool

Most readers can't name what's being done to them while they're reading it. Modern news and social posts use a wide vocabulary of subtle persuasion techniques — historical revisionism, dehumanization, false equivalence, manufactured outrage. This tool is a media-literacy aid: it shows you, on real text you encounter, exactly which technique appears where, and why.

1
Paste a paragraph from an article, tweet, or post you find suspicious.
2
In about half a second you see a verdict, a high-level intent in its color, the specific technique, and the exact phrases that triggered it.
3
Hover any highlight for the technique name, a one-sentence definition, and the model's confidence.
4
Over time you internalize the vocabulary — and start spotting the patterns in your feed without the tool.

The entity-framing layer adds the who: beyond "this text is propagandistic," it shows which actors are cast as protagonist, antagonist, or innocent — and the specific role, like Guardian, Tyrant, or Victim. That reveals the rhetorical narrative architecture, not just an inventory of techniques.

What this is not:

A fact-checker — it does not verify claims against reality.
A content-moderation or censorship aid — the license restricts mass-screening use.
A definitive verdict — it is statistical; low-confidence predictions are de-emphasized on purpose.

Languages supported: English, German, French, Italian, Romanian, Russian, and Ukrainian, on a multilingual mBERT / XLM-R backbone. There's also a Chrome extension that sends the same request from any page you're reading.

Participate in Our Research

You are invited to participate in a research study conducted at the Quality and Usability Lab, TU Berlin. This study investigates how users interact with and evaluate a website for identifying propaganda in text. We are particularly interested in how users perceive the system's explanations and whether they help users understand the system's assessments.

What participation involves

1
Answer a few short questions about yourself, how helpful and trustworthy you found the explanations, and whether you agree with the system's analysis.
2
Use the propaganda detection tool embedded in this page (Section 3 of the user study) to analyze at least 3 text passages of your choice.
3
This takes approximately 5-10 minutes.

Your rights & data handling

  • Participation is voluntary. You may stop participating at any time by closing this page without giving a reason.
  • The study records your interactions with the tool and your questionnaire responses. No directly identifying personal information will be collected.
  • To help ensure data quality, the study stores a randomly generated pseudonymous participant identifier in your browser and submits it together with your responses. This identifier is used to detect duplicate submissions and cannot directly identify you.
  • If you would like your submitted data removed after participation, please contact .
  • Please note that anonymized responses may be used in academic publications and shared with collaborators.

Research Team

This study is conducted by the Quality and Usability Lab at TU Berlin.

For questions about the study, please contact:

  • Aaron Eidt ()
  • Mara Hendel ()
  • Moritz Falk ()
  • Mykyta Zabashtanskyi ()
  • Anna Stiegler ()

Supervisor: Veronika Solopova ()

Consent

By clicking I agree and want to participate below, you confirm that you have read and understood this information and voluntarily consent to participate.