How a claim check works

Not one opinion.
Evidence, then judgment.

Claim identified Evidence Engine AI checkers Judge Result

1. Find the claim

Sherlock finds the exact statement to check.

A post can mix a headline, caption, date, opinion and several factual claims. Sherlock separates them and identifies the checkable statement.

If there is more than one possible claim, TruSight asks you which one you want investigated.

2. Evidence Engine

Then we find the evidence worth using.

The Evidence Engine searches broadly, but it does not choose sources because they agree with the claim. Relevance comes first: does this source address the same people, action, place and time?

It then grades who published it, what kind of evidence it contains and whether it is current. Repeated copies do not become separate confirmation. Only the strongest on-claim sources go forward—and if none qualify, the evidence set can be empty.

1. Match the claim

The source must address the actual event or assertion—not merely mention the same topic.

2. Grade the source

Origin and evidence type matter. A direct record carries different weight from a repost or opinion.

3. Remove false agreement

Repeated publishers, copied reports and outdated material do not count as independent confirmation.

3. Tribunal

The AI checkers test both sides.

The checkers do not search the web and they do not choose the evidence. Both receive the same evidence selected by the Evidence Engine.

One builds the strongest honest case that the claim holds. The other tests whether it is false, misleading, incomplete or missing context. Each must ground its argument in the selected sources and acknowledge weaknesses.

AI checker · Support Builds the strongest honest case for the claim, cites the selected evidence and names weaknesses in its own case.
Judge Reads both arguments and the same evidence. It does not average their views; it sets the final claim score and records uncertainty.
AI checker · Challenge Tests contradictions, missing context and alternative explanations, using the same selected evidence.

4. Result

The answer comes with its working.

The final result shows the claim score, a plain-language finding, the evidence that was selected, both checker arguments and what could not be confirmed.

  • Finding
  • Selected evidence
  • Checker arguments
  • What remains uncertain

Scroll inside the phone to inspect the full result.

TruSight
34
Claim score Likely misleading

Finding

The claimed nationwide subsidy is not supported by the selected official records and reporting. The post overstates a limited regional programme.

Checked as of 16 Sep 2026

Claim checked

Every household can register online for free electricity under a new nationwide scheme.

Evidence selected

How it was checked

AI checker · Support
Case for the claim

A regional utility does offer limited bill relief, so the post contains a small element of truth.

AI checker · Challenge
Case against the claim

The selected records do not show a nationwide programme, and the registration link is not an official government domain.

Judge
34

Weighs both arguments against the selected evidence and sets the final score.

What remains uncertain

Local eligibility can differ by state and utility. No selected source confirms the nationwide offer shown in the post.

See the evidence.
Understand the result.

Start a Check