Spot the rhetoric before it shapes your view

PropaLens underlines sentences that use one of 14 known persuasion techniques, such as or , right in the article, with the definition one hover away. Each highlight is a signal to pause and check a claim. Free, and your reading never leaves your device.

Add to Chrome See how it works Try a sample article

Free. Experimental: the model makes mistakes, so treat highlights as a prompt to read more closely.

Why it works

Persuasive techniques lose much of their power the moment you recognize them. PropaLens flags a technique while you're reading — when it can still change your mind about how you feel — instead of leaving it invisible. Over time, the hover-card definitions act like spaced practice: you learn the techniques and start spotting them on your own, with or without the extension.

What it does

Underlines in the page

Article text underlined in different colours to mark detected propaganda techniques

Each flagged sentence gets a coloured underline. The colour shows the technique.

Short explanation on hover

Hover or tap an underline to see the technique name, a one-line definition and a Low, Medium or High confidence level.

Page summary

Hover card explaining appeal to fear or prejudice, with a confidence label and definition

The popup lists the techniques found with counts. You can hide a technique or jump to the next example.

Adjustable strictness

PropaLens settings with the strictness slider set to 50 percent and model options

A slider trades coverage for reliability: fewer highlights that are more often correct, or more highlights that include more false alarms.

Local. Models are bundled with the extension and run in the browser. No page text is uploaded.
Opt-in. Nothing is scanned until you click "Analyse this page" or enable a site.
Multilingual options. Several model choices, including multilingual ones, are available in settings.

How well does it work?

We evaluated the three bundled models on 2,087 held-out sentences derived from the PTC corpus. The original corpus labels text spans; for this evaluation, we assigned those span annotations to sentences. The precision/recall chart shows how often highlights are correct and how much labelled propaganda is found at each strictness setting. The F1 chart combines those measures into one score. A higher setting gives fewer highlights that are more often right. Results on other kinds of text will differ.

Line charts of precision and recall by strictness setting for MiniLM and XLM-RoBERTa. — Precision and recall by strictness setting (50% is the default). At the default setting, precision is about 58–60% and recall is about 76–77%. Measured on an English validation set: real-world performance, especially in other languages, may vary.

Training data and method

The models are fine-tuned classifiers built from two public datasets. We changed the data as follows:

SemEval-2020 Task 11 annotates text spans. We converted these to sentence-level examples with one technique per sentence.
PropXplain texts were mapped to the SemEval technique classes with a language model, and we checked a sample by hand.
Classes with too few examples were extended with synthetic examples generated by a larger generative language model.

SemEval-2020 Task 11 (PTC corpus): Da San Martino, Barrón-Cedeño, Wachsmuth, Petrov and Nakov, 2020. Dataset, paper. Licensed CC BY 4.0.
PropXplain: Alam et al., Findings of EMNLP 2025. Dataset, paper. MIT License.
We modified the data as described above. This project is not endorsed by the dataset authors.

Planned

PropaLens Cloud

We are considering an optional hosted service that uses a larger model and explains why a sentence was flagged. It would cost roughly $1 to $3 per month; the price and features are not final. The local version stays free. Leave your email if you want to hear when there is something to try. We will only use it for this.

We only use your email to tell you about PropaLens Cloud. See the .