Truthlume isn't a lie detector. It helps you reach a better-founded conclusion by separating what can be checked against sources from what can't.
The process
- 1We read the content you submit. Links to news sites are fetched; screenshots are read with an image-reading AI model.
- 2An AI model picks out up to five of the most important factual claims. Opinions and value judgments are listed separately.
- 3For each claim we look for existing fact-checks (Google Fact Check Tools) and run a web search, prioritising original sources.
- 4The AI assigns a category only based on the sources found. If there aren't enough sources, the claim is marked as "Can't be verified yet".
What each category means
Reliable sources found confirm the claim.
Reliable sources found say the claim is not accurate.
The claim is based on something real but is presented in a way that distorts it, or omits important context.
We couldn't find enough reliable sources to support or contradict it. This doesn't mean it's false.
Confidence levels
High means several independent, reliable sources agree. Medium means sources exist but are limited or partly indirect. Low means the evidence is thin.
Where sources come from
Published fact-checks from organisations indexed by Google Fact Check Tools, and web results prioritising official bodies, public statistics, original documents, established media and fact-checkers. Every link shown comes from these searches — the AI is not allowed to invent sources.
Limitations
- AI can misread content or misjudge sources. Always open the links.
- Very recent events may not be covered by sources yet.
- Images and videos without context are hard to verify.
- Some social networks can't be read from a link; paste the text instead.
- We analyse at most five claims per piece of content.
Privacy
No accounts. Submitted content is deleted automatically after 30 days. To control costs we limit each connection to 10 analyses per day.