CleanFeed: Verified, Transparent Journalism in An Era of Disinfo and AI Slop
,Bad content is flooding the internet – and fact-checking alone can't fix it. As part of the newly launched CleanFeed research project, we're working on a different approach: building transparency and media literacy directly into journalism itself, so audiences can see for themselves why a piece of content is worth trusting.
The internet is drowning in digital garbage. AI-generated slop, manipulated media, coordinated disinformation – low-quality and outright false content is no longer the exception. In many corners of the web, it's become the norm. And the uncomfortable truth is: No verification operation in the world, however well-staffed and well-funded, can keep up. The current information ecosystem produces bad content faster than anyone can analyze and debunk it.
That's why we're increasingly convinced that the answer isn't just more and better fact-checking ex post. It's building transparency into the content itself, from the very beginning – so that audiences don't have to take trustworthiness on faith, but can actually see it.
That's the core idea behind CleanFeed, a new research project funded by the German Federal Ministry for Research, Technology and Space (BMFTR), running from March 2026 through March 2029. Together with our project partners – Fraunhofer FOKUS (who lead the consortium), castLabs, and Geißendörfer & Leschinsky (G&L) – we're working on solutions that allow audiences to make genuinely informed trust decisions about the media they consume.
Provenance as Prevention
CleanFeed is built around a simple but powerful premise: If people can see where a piece of content comes from and how it's been processed, they're in a much stronger position to judge whether to trust it. Rather than reacting to disinformation once it's already circulating, the project focuses on structural transparency as a preventive measure.
A central technical pillar here is C2PA, the standard developed by the Coalition for Content Provenance and Authenticity. It's still evolving, and has a lot of rough edges, but our colleagues at DW IT Projects have been deep in this space for a while now. We're also looking at alternative and complementary approaches like soft-binding watermarks or the ISCC (International Standard Content Code). The idea is to keep our options open.
What We're Building
Within the broader project, our team at DW Innovation is developing a modular feed system designed specifically for journalistic content. But we want to be clear: this isn't just about stamping articles with a digital certificate and calling it a day. The technical layer–provenance markers, content certificates, open protocols like ActivityPub – is a starting point, not the goal.
What we aim for is giving audiences a clear perspective on how a piece of journalism was made. What sources does it draw on? How is it framed? What editorial decisions shaped it? That kind of transparency doesn't just help people spot disinformation. It builds the media literacy that makes audiences more resilient over time.
Our feed system brings together three interlocking concepts to do exactly that:
CleanFeed. A clean, filterable timeline of certified content, built on content certificates and open protocols, a curated stream you can trace back to its origin. As a news consumer, you'd know everything here has been verified and sourced.
No need to look for specific features all the time; this is a place you can trust.
CheckAgain. Developed in partnership with with DW's editorial teams, this label indicates that content has been reviewed multiple times and verified as current and accurate. It's a lightweight but meaningful signal that the journalism holds up to scrutiny.
DeepDive. This is where we go beyond the certificate. DeepDive offers contextual analysis, background material, and critical framing around a piece of content. It's designed to invite readers not just to consume journalism, but to interrogate it – to understand the perspective behind a story, the evidence it rests on, and where its limits are. In other words: Media and information literacy built right into the reading experience.
Who's Involved – and What We Do
DW's role in CleanFeed spans three areas: requirements analysis and focus groups (making sure what we build actually reflects how real users think about trust and transparency), overall architecture and design (shaping the system's structure and specifications), and science communication and knowledge transfer (ensuring the project's findings don't stay locked inside a research paper).
Our partners bring complementary and highly specialized expertise to the table. Fraunhofer FOKUS, one of Germany's leading applied research institutes in the field of open communications systems, leads the overall project. castLabs, well known in the media tech industry for digital rights management (DRM), forensic watermarking, video players, and content processing, brings deep technical know-how around content security and distribution. Geißendörfer & Leschinsky rounds out the consortium as an experienced IT services provider and system integrator with a strong focus on audio and video streaming solutions.
CleanFeed is part of a wider family of projects at DW Innovation: its sister projects Fake-O-Meter (AI-assisted verification) and PADSE (audio forensics) tackle overlapping problems from different angles. Together, they reflect a shared conviction: In the fight against disinformation, there's no single silver bullet. It takes verification and analysis, de-bunking and pre-bunking, provenance technology and media literacy, cutting-edge infrastructure and old-fashioned editorial judgment. They're vital components in the process of (re-)building something that's harder to bring forth than any deepfake: trust.
