Public research dataset
Clarity Evidence Index
A rolling 90-day window into the evidence graph behind Clarity Today: public articles, the article claims readers can inspect, and the stored evidence relationships connected to those claims.
This describes what Clarity has published. It is not an objective truth score, a source rating, or a newsroom target.
5,083
Published articles
98,915
Article claims
105,410
Evidence relationships
Source data through
Relationship mix
Each relationship records whether stored evidence supports, contradicts, or contextualizes a published article claim. A contradiction records a material conflict; it does not decide which source is correct.
87,382
Supports
1,705
Contradicts
16,323
Contextualizes
Daily record
| Date | Published articles | Article claims | Evidence relationships | Supports | Contradicts | Contextualizes |
|---|---|---|---|---|---|---|
| 2026-08-12 | 19 | 451 | 0 | 0 | 0 | 0 |
| 2026-08-13 | 502 | 10,232 | 0 | 0 | 0 | 0 |
| 2026-08-14 | 453 | 8,823 |
The six desks
| Section | Published articles | Article claims | Evidence relationships | Supports | Contradicts | Contextualizes |
|---|---|---|---|---|---|---|
| Security | 453 | 8,352 | 7,582 | 6,508 | 99 | 975 |
| Build | 1,524 | 30,332 | 31,734 | 26,893 | 398 | 4,443 |
| Invest | 1,244 | 24,544 | 27,118 |
Methodology
Window and cadence
The inclusive window covers 2026-05-26 through 2026-08-23 using publication dates in America/Los_Angeles. It is updated from the source of truth at most once an hour; a cache refresh does not change the modification time.
Exact measures
- Published articles
- Distinct public articles that cleared publication in the window.
- Article claims
- Distinct article-claim records attached to those public articles.
- Evidence relationships
- Distinct article-claim and stored evidence pairs, classified by their recorded relation.
Limits
- Counts describe Clarity’s published evidence graph, not an objective truth score.
- A contradiction records a material conflict and does not decide which source is correct.
- Publisher count and source quality are intentionally absent because current identity coverage is not defensible as a headline measure.
- Late corroboration, claim corrections, and evidence attachment can revise a prior day.
- The oldest day leaves the window when a new Pacific publication day begins, so these are not lifetime totals.
- Search visibility and AI citations are measured externally in search-provider reports, never inferred here.
Read the broader publication methodology.
Download the data
Both distributions come from the same validated aggregate as this page.