How it works
Skepticle checks the key factual claims in an article against published sources and shows you exactly where it looked. It is a starting point for your own judgement, not a verdict.
How is the score made?
We pick out up to five key factual claims. For each one we gather trusted sources and read what each says: does it confirm the claim, dispute it, or not say either way? The claim’s score is the share of sources that confirm it. If five trusted sources confirm a claim and one disputes it, the reliability score is 83%. The overall score combines the claim scores, giving more weight to the claims that matter most to the article.
Which sources do you use?
In order of trust: fact-checks published by established fact-checking organisations; official sources such as governments and international bodies; peer-reviewed research and recognised medical institutions for health and science claims; and reports from established news organisations, found with a live web search that looks at trusted sites first. For topics those sources don't cover, we may use specialist sources, such as the official site of the company or organisation a claim is about, labelled with the reason; on their own they can give at most “Mostly true” or “Mostly false”. Other websites and Wikipedia (which anyone can edit) are shown only as context and never counted. Satire and sites known for publishing false information are never used. Every source is labelled by type and linked under the claim, so you can read it yourself.
How are sources counted?
Each organisation gets one vote, so several pages from the same outlet count once. Only trusted sources that take a clear side are counted; those that are merely related are shown as context. With a single source we don’t show 100 or 0: the rating stays cautious (“Mostly true” or “Mostly false”) and says so. If no trusted source takes a side, we mark the claim Unverifiable rather than guess. Under every claim you can see the count and which side each source is on.
What is "AI's best guess"?
When no reliable source settles a claim, we don't give it a reliability score. Instead we may show the AI's best guess, clearly labelled as not confirmed by sources. Treat it as a hint, not a fact-check.
Will I get the same answer twice?
The reliability score is a count of sources, not a number the AI picks, so the same sources give the same result. The same link or text checked again within 24 hours returns the saved result. For breaking news, results can change as more trusted sources publish.
What does the reliability score mean?
It tells you how strongly trusted sources back the checkable claims: 100% means every source we found confirms them, 0% means every source disputes them. 65–100% is reliable, 40–64% is mixed, 0–39% is unreliable. “Unverifiable” means we could not find enough to judge. It measures agreement among sources, so one outdated or dissenting article can pull a reliability score below 100%.
What doesn't it mean?
The reliability score is not a rating of the author, the publication or its politics. It doesn't judge opinions, and a low reliability score doesn't mean every sentence is wrong.
Can the AI make mistakes?
Yes. It can misread a claim or miss a newer source. That is why every result shows its sources. Please review them before sharing a result.
Are you neutral?
We apply the same process to every article, whatever its topic or point of view. We do not take sides, accept payment for results or adjust scores by hand.
What happens to what I paste?
We use it to run the check and create a shareable result link. See the Privacy page for details.