What is source attribution and how does NO_ATTRIBUTION reduce AI citability?

Source attribution is the practice of naming the origin of a numeric or factual claim within the same passage that contains the claim. In AEO content, this matters because AI engines use attribution phrases as trust signals when deciding whether a passage is authoritative enough to relay to a user. NO_ATTRIBUTION fires in Seeklyn when a passage contains at least 1 fact token (a digit, year, percentage, or dollar amount) but contains none of the ~12 recognized attribution patterns.

The severity of NO_ATTRIBUTION is low — it is the only one of Seeklyn's 8 flags with low severity. The evidence grade is moderate: correlational data consistently shows that attributed claims are cited more often, but the relationship is less structurally deterministic than a broken anaphora or missing answer-first opener. A passage that fires NO_ATTRIBUTION is still extractable; it just carries a weaker authoritativeness signal.

Seeklyn's `isAttributed` function runs a regex over the full passage text and returns true if any of the following patterns appear: "according to," "reported by," "data from," "survey by," "cited by," "research by," "study by/in/of/from/published," "published in," "researchers," "scientists," "et al," "journal of," "per," or a standalone 4-digit year (matching \d{4}). The check is case-insensitive and matches anywhere in the passage.

NO_ATTRIBUTION fires only when two conditions are both true: `hasFactToken(chunk.Text)` and `!isAttributed(chunk.Text)`. A passage with no numeric data at all does not fire NO_ATTRIBUTION — but it will fire LOW_FACT_DENSITY. A passage with "87%" and "per HubSpot 2024" satisfies both conditions and clears both flags simultaneously.

The minimum change to clear NO_ATTRIBUTION is appending a 4-digit year to the sentence containing the figure: "Video content increases engagement by 87% (2024)." A more robust attribution names the source: "Video content increases engagement by 87% compared to text-only posts, according to HubSpot's 2024 State of Marketing report." Both forms satisfy Seeklyn's attributionPattern regex.

For maximum citability, name the source, include the year, and place the attribution clause in the same sentence as the figure — not in a separate footnote or end-of-paragraph aside. AI engines chunk text at paragraph boundaries; attribution in a different paragraph than the claim may be separated during retrieval, causing the chunk to fail the check even if the page nominally cites its sources.