The Reality Gap
This project has been decommissioned
BiasFeed stopped publishing in September 2026. The automated pipeline that produced these analyses has been switched off, and no new articles will be added.
It ran unattended for ten months and analysed 583 articles. It was built to settle a question, described below, and it settled it. Everything it produced stays online and free to read.
This project didn't start with a business plan. It started with an argument.
I realised that my friends and I weren't just disagreeing on opinions; we were living in different realities. We were reading the same headlines but seeing completely different stories. Media bias isn't just annoying. It's an invisible filter that shapes how we perceive the world.
BiasFeed is not here to tell you what to think.
It attempts to show how you are being told to think.
We don't believe in "Unbiased News". It doesn't exist. Every writer makes choices. Instead, we believe in Radical Comparison. By using AI to strip a story down to its raw facts and then deliberately spinning it Left and Right, we make the invisible mechanics of bias visible.
Frequently Asked Questions
What is BiasFeed?
BiasFeed is a media literacy tool that takes real news articles, strips them down to their core facts, and then rebuilds them through three different editorial lenses: progressive, conservative, and centrist. The goal is to make the mechanics of media bias visible so readers can develop sharper critical thinking skills.
How does BiasFeed work?
Each article goes through a three-step process. First, an AI model reads the source article and extracts only the verifiable facts, removing all opinion, framing, and charged language. Then, three separate AI personas rewrite the story from different political perspectives. Finally, readers can compare all versions side by side and use a bias highlight toggle to see exactly where the framing differs.
Is BiasFeed biased itself?
We do not claim to be unbiased. What we claim is transparency. By deliberately showing how the same facts can be spun in different directions, we make bias visible rather than pretending it does not exist. The fact extraction step uses an AI model that may carry its own subtle biases, which is why we always link to the original source article for verification.
How are articles selected?
Articles are selected from major international news outlets covering politics, policy, conflict, economics, and social issues. Two new analyses were published daily while the project was running. Selection prioritises stories that are widely covered across the political spectrum, where framing differences are most revealing.
Is BiasFeed still running?
No. BiasFeed was decommissioned in September 2026 and is no longer being updated. The archive of 583 analysed articles remains online and readable, but the pipeline that generated them has been switched off.
What AI models does BiasFeed use?
BiasFeed uses large language models for fact extraction and perspective rewriting. The specific models may change as the technology evolves, but the methodology stays the same: extract facts first, then rebuild narratives separately. The AI is a tool in the process, not the editorial voice.
Don't trust us. Trust the comparison.