How Trace works — and what it won't do

Fact-checkers don't stare at a single page trying to decide if it's trustworthy. They leave it. They open new tabs and ask what other, independent sources say about the same claim. That habit is called lateral readingwhy?.

Trace automates the first move: you paste a claim, and it gathers real sources that mention it so you can compare them side by side. The reading — and the judgment — stay with you. The point is for the habit to become yours.

Trace never prints "true" or "false." It lays out provenance: who said it, where it first appeared, and how each retelling changed the words. A small hedged preprint becoming a viral "proven" headline is a pattern you can see for yourself — no scoring, no editorial verdict.

That stance is deliberate. A confidence score invites you to outsource your judgment to a number. A trail invites you to look.

• Brave Search. Live lookups run against the Brave web index for sources that mention your claim. The search key stays on the server — your query is sent to Brave to fetch results, and nothing about you is stored by Trace.
• Web archives. Once you have the sources, the original and its earliest copies are what matter most. Reading them in order — and checking archived versions when a page has quietly changed — is how you find where a claim actually started.
• Seed examples. When no live key is configured, Trace shows worked examples so the method is still teachable. They are clearly labeled as examples, never as live results.

Every trail says where it came from. Live results are labeled as live and listed by host so you can open the originals. Seed examples are labeled as examples. Trace will never tell you that you are "safe" or that a claim is settled — it shows you the trail and trusts you to read it.

Trace is free, built by America's Future, and currently in beta — the method is solid; the engine is still being refined.