State of climate misinformation and disinformation in audiovisual media in 2025

The dataset contains files related to climate disinformation and misinformation published in open-data on the Ecology Media Observatory website. They are the result of a collaboration between the [Observatoire des Médias sur l’Ecologie](https://observatorymediaecologie.fr/) and [Science Feedback](https://science.feedback.org/fr/), a leading fact-checking organisation in France on climate issues, as part of the Climate Safegards project. The whole process was also produced with the support of our expert committee. These data make it possible to take stock of the evolution of climate misinformation in the audiovisual sector since January 2025, to study the distribution of cases identified by fact-checkers between different media, and to observe the different narratives of climate disinformation prevailing. #### **In particular, it contains the following files** * evolution\_of\_misinformation.csv * misinformation \_per \_media * disinformation \_narratives \_with \_debunks #### Methodological Preamble In **academic literature**, climate disinformation is generally defined as follows: \*\*Climate disinformation \*\*is defined as a false or misleading discourse with a high risk of misleading the public on facts proven by the state of scientific knowledge about climate change and climate action regarding mitigation and adaptation measures as established by the IPCC.  * **Climate misinformation** is distinguished by the speaker’s lack of demonstrated will to harm, which may therefore be error, or permeability to misleading narratives. These **works adopt an operational approach**, which focuses mainly on:  * the false nature of the contents,  * their potential negative impact on audiences or public policies, rather than on the intentionality or awareness of producers and broadcasters. In this context, two additional terms are used to refine the analysis: * **False affirmation**: unsubstantiated claim, either scientifically contradicted, manipulative by omission, or based on invalidated theories (see below). A statement contradicted in the sequence (2 minutes) is not counted (see 4. 100% human, ethical and robust validation). * **Narrative of disinformation**: Among detected misinformation cases, a recurring narrative emerges significantly if more than > 8 occurrences are detected. Repetition is considered strong enough to indicate the likely existence of a certain intentionality to mislead the public. #### Dataset processing and compilation The data shall be compiled in accordance with the following steps: * Acquisition of raw data (audiovisual transcription of TV/Radio streams) by API * Identification of sequences dealing with climate change (methodology based on thematic keyword dictionaries) * Detection of cases at risk of containing climate misinformation (fine-tuned AI model on data labeled by hand by fact-checkers) * Manual fact-checking of all detected cases to ensure the robustness of the data produced * Association of cases with recurrent narratives of disinformation (if > 8 cases corresponding to the same narrative) #### Update and access These datasets will be updated every 3 to 6 months, according to the resources available for the Ecology Media Observatory. These datasets were built according to a fully open methodology, which users can [see here](https://observatorymediaecologie.fr/methodologie-mésinformation-climatique/).