How Caus reads cited evidence
Caus compares the wording of the sources cited for a claim and groups the ones that repeat the same story. What remains is a clearer view of how many distinct stories the citations really represent, plus a short report explaining each grouping decision. This source-independence analysis is the one check Caus performs today.
It does not fetch URLs, judge whether the claim is true, or score how strong the evidence is.
What source independence means
Citations are not the same as independent sources. Several articles can repeat the same original reporting, the same quote, the same press release, or the same syndicated wire copy. Counting them as separate backing makes a claim look better supported than it is.
A source can be useful even if it is not independent. Caus simply makes repeated coverage visible so one echoed story does not look like many separate confirmations.
- Caus looks for overlap in the wording of the cited excerpts to tell repeated coverage apart from more independent coverage.
- This is a structural read of the citations — it reflects how the sources relate, not whether the claim is true.
How echo grouping works
Caus uses conservative text-overlap checks on the excerpts you provide. If two cited excerpts substantially repeat the same wording or coverage, Caus groups them as repeated coverage. If an excerpt does not closely overlap with others, it is left with no strong match.
- 1You paste an AI answer, article, research summary, or source list, then add the cited source text.
- 2Caus compares the text of those excerpts to each other.
- 3Strongly overlapping excerpts are grouped as repeated coverage; excerpts that do not closely overlap are left with no strong match.
The public mode is deliberately conservative: near-verbatim repeats are grouped, but paraphrased echoes or lightly rewritten coverage that share meaning without sharing wording may stay separate. A citation left with no strong match is not a certificate of independence — it means Caus did not detect obvious repeated coverage, not that the source is confirmed independent. Caus only reads the excerpts you provide — it does not open links, fetch URLs, or look for sources you did not include.
What a citation report contains
The report explains which citations were grouped, which had no strong match, and why the grouping happened. It is meant to make citation structure easier to inspect, not to decide whether the claim is true.
Every report shows
- How many citations were reviewed in total.
- Which citations were grouped together as repeated coverage into one story.
- Which citations had no strong match with any other citation.
- Why each grouping happened — the overlap Caus found between the cited excerpts.
- Overlap notes for grouped citations, such as the strongest and average similarity between their excerpts.
- A plain note on what was not checked — the report explains structure, not whether the claim is true.
Illustrative report
SyntheticClaim: “A city pilot program expanded overnight bus service.”
3 citations → 2 distinct stories
City transit announcement + Local newspaper article
Why grouped: the excerpts repeat the same announcement language about the pilot scope and launch date.
University mobility lab analysis
Why no strong match: the excerpt discusses separate ridership context rather than repeating the announcement wording.
This report shows citation structure. It does not decide whether the claim is true.
What Caus does not do
Source independence is one structural read of the citations behind a claim. To keep that read honest, here is what Caus deliberately does not do today.
Caus does not
- Decide whether the claim is true.
- Fact-check claims or check them against the outside world.
- Say whether a source supports or contradicts the claim.
- Fetch URLs, open links, or read anything beyond the excerpts you paste.
- Search for or discover sources you did not provide.
- Score how strong, accurate, or trustworthy the evidence is.
- Use an AI model or large language model to review your sources.
See it on your own sources
Paste an AI answer, article, or source list and read your citation report — how many distinct stories the citations represent, and which repeat the same coverage.