We use multiple AI models to extract claims, identify reasoning patterns, and compare how different sources cover the same story. Here's how it works.
We Analyze Reasoning, Not Truth
We don't tell you what's true. We show you how arguments are constructed.
A claim can be factually true and still be logically manipulative. A politician can cite real statistics while committing a logical leap. A journalist can report accurate facts while framing them to favor one side.
Our AI identifies these patterns. It flags when evidence is missing, when only two options are presented as if no others exist, when emotional language substitutes for argument. These aren't verdicts—they're prompts for your own critical thinking.
The same standard applies to everyone. Left, right, center—any side can have flawed reasoning. Any side can be logically sound. We enforce rigor, not ideology.
What the AI Actually Does
Extracts Who Said What
AI identifies speakers, their claims, and the evidence they cite. It separates what sources actually said from how journalists frame it—so you can see editorial choices clearly.
Flags Reasoning Patterns
Using a fixed, transparent taxonomy of logical issues, AI identifies patterns like missing evidence, false dichotomies, or oversimplification. The language is intentionally neutral: "Needs More Evidence" not "Lying." These are observations, not accusations.
Compares Perspectives
AI analyzes how left, center, and right sources cover the same story. It surfaces what everyone agrees on, what's disputed, and—most importantly—the underlying value tension driving the disagreement.
Answers Your Questions
The chat assistant lets you dig deeper into any claim. Ask about evidence, context, or how different sides see an issue. The AI explains perspectives charitably without advocating for them.
Multiple AI Models, Your Choice
Different AI models have different training data and tendencies. A model trained primarily on Silicon Valley data may have different blind spots than one trained elsewhere.
We run analysis through multiple models to reduce single-point-of-bias. In the chat interface, you choose which model you talk to—including Grok for those skeptical of mainstream tech platforms.
You always know which model you're interacting with. No hidden switches.
Prompts Designed for Neutrality
Every prompt instructs AI to be a neutral observer, not an advocate. When explaining a conservative perspective, the AI doesn't mock it. When explaining a liberal perspective, the AI doesn't endorse it.
Chat responses are evaluated against explicit criteria:
- Non-partisan tone — No loaded language, no one-sided framing
- Charitable representation — Explain why someone might hold a view, even if you disagree
- Explain, don't advocate — "This perspective holds that..." not "The truth is..."
- No partisan sources — We don't cite Fox News, MSNBC, Breitbart, or HuffPost in responses
What AI Gets Wrong
We're not going to pretend AI is perfect. Here's where it falls short:
- It can misidentify speakers or misattribute claims
- Logical issue detection isn't perfect—sometimes it flags things that aren't issues, sometimes it misses things that are
- It doesn't fact-check—it analyzes argument structure, not whether claims are true
- Training data has cutoff dates—it may not know recent events
That's why we always link to the original sources. Check our work.
Verify Everything
Don't trust us. Verify.
- Read the sources — Every claim links to the original article
- Switch models — If you're skeptical of one AI provider, try another
- Compare perspectives — See how left and right sources frame the same facts
- Flag issues — If you spot something wrong, tell us
The goal isn't to tell you what to think. It's to give you better tools to think for yourself.