Methodology
TruthBubble is being built around source-backed context, visible uncertainty, and careful verdicts for user-initiated verification.
TruthBubble is designed to focus on the factual claim being made instead of scoring an entire post, creator, or platform account. A caption, screenshot, reel, link, or forwarded message can contain multiple claims, and each may need different evidence.
Direct evidence can support or contradict a claim clearly. Indirect evidence may only provide context, timing, background, or related facts. TruthBubble should avoid treating indirect context as proof when the claim needs stronger support.
Official records, primary documents, reliable reporting, expert statements, and independent corroboration are treated differently. When sources disagree, the result should explain the disagreement rather than hide uncertainty.
Images and video frames can show what appears in the content, but visibility is not always proof. Screenshots may omit context, captions can misdescribe old footage, and manipulated media can be difficult to identify from a single frame.
Confidence is not certainty. It is a signal about the available evidence, source agreement, claim specificity, and whether context is missing. Results can change when better evidence appears.
Fast-moving stories can change. TruthBubble should support correction-minded language, encourage source review, and avoid framing an early result as final when new evidence is likely.
TruthBubble assists verification. It does not replace professional fact-checkers, journalists, or qualified experts. For important medical, legal, financial, safety, or civic decisions, review original sources and seek appropriate expertise. Read more about AI verification limitations.