AI against disinformation · 2026–2029
A multimodal AI companion that helps people check disinformation — and strengthens democratic resilience.
We bring deepfake detection, fact-checked explanations and a conversational assistant into one citizen-facing tool, tested in the field around real elections. It is built to support human judgement — not to act as an automated arbiter of truth.
⬡ Funded by the German Federal Ministry of Research, Technology and Space (BMFTR)
About
Fake-O-Meter develops a multimodal, AI-based assistance system that helps people identify and understand disinformation. It combines deepfake detection across image, audio and video, explanations grounded in professional fact-checks, an adaptive dialogue assistant, and pre-bunking campaigns around elections.
Crucially, it is designed as a companion for human judgement — surfacing evidence, explaining reasoning and pointing to trusted sources — rather than handing down automated verdicts. Fact-checkers and editors stay in the loop throughout.
Approach
The “Deconstructor” analyses image, audio and video for signs of manipulation, and explains what it finds.
MultimodalityRetrieval-augmented answers grounded in professional fact-checks from our media partners.
RAG & explanationAn adaptive dialogue that explains and discusses, attuned to context, language and audience.
DialogueShort, evidence-based campaigns that prepare people for disinformation ahead of elections.
ResilienceTimeline
Each election is a real-world test: the assistant and pre-bunking campaigns are deployed, then evaluated with citizens.
Consortium
German Research Center for Artificial Intelligence — leads integration, the public demonstrator and the overall coordination of the project.
Quality & Usability Lab — retrieval-augmented explanation, and fairness, trust & transparency.
Visit site ↗Otto von Guericke University — dialogue, multimodality and the user interface.
Visit site ↗Ernst-Abbe-Hochschule Jena — the assistant concept and structured, fact-based discussion.
Visit site ↗Evaluation, fact-checking expertise and editorial use cases.
Visit site ↗Deutsche Presse-Agentur — requirements, ELSI, editorial use cases and pre-bunking.
Visit site ↗The demonstrator platform: deepfake detection and the conversational fact-check companion.
Visit site ↗Coordinated by DFKI · scientific lead Prof. Dr. Sebastian Möller.
Funding
Fake-O-Meter is funded by the German Federal Ministry of Research, Technology and Space (BMFTR) within its research programme on IT security and communication systems. Project term: May 2026 – February 2029.
Project profile on the BMFTR research portal ↗Contact
Fake-O-Meter is coordinated by DFKI’s Speech & Language Technology group. For questions about the project, collaboration or press, reach the coordination team.