Citation Trail
Source Mapping for Generated Claims
Hover any sentence in an AI answer and the contributing sources highlight in place. Each fact carries a visible thread back to where it came from — no more anonymous synthesis behind a single footnote number.
Framing
The problem
Sourced AI hides which fact came from which source behind aggregate footnotes.
The pattern
Map every claim to its contributing source spans; reveal the link on hover.
Why chat breaks here
Chat citations are flat numbers — readers cannot verify which sentence is grounded where.
Risks
False precision: a visible source link can imply rigor that the underlying retrieval lacks.
Avoid when
The output is opinion or synthesis where granular source mapping would be misleading.
Use when
Readers must verify which source supports which specific claim, not just the document overall.
DOPE evaluation
- Directability
- Click a citation to inspect or remove its source from the answer
- Observability
- Every claim shows its source spans on hover
- Predictability
- The same claim reliably maps to the same set of source spans
- Explainability
- Each output sentence is bound to specific source spans, not just a footnote number
In the wild
- Perplexity (Perplexity) — Inline numbered citations linked to a side rail; hover preview of cited source. The canonical example of the pattern, even though source-to-claim mapping quality is contested.
- Glean (Glean) — Deep-linked citations jump to the exact passage that supports the statement; hover preview highlights the cited text. One of the strongest enterprise examples.
- Notion AI Q&A (Notion) — Q&A returns answers with inline citations linking back to source pages and connected apps. Per-claim mapping is weaker than Perplexity but valid as a workspace-grounded example.
- Granola 2.0 (Granola) — Inline citations + jump-to-source on every claim in the meeting summary, mapping to transcript timestamps. A 2026 addition that proves the pattern travels beyond search.
FAQ
When should I use the Citation Trail pattern?
Readers must verify which source supports which specific claim, not just the document overall.
When should I avoid the Citation Trail pattern?
The output is opinion or synthesis where granular source mapping would be misleading.
What problem does Citation Trail solve?
Sourced AI hides which fact came from which source behind aggregate footnotes.
Why is chat the wrong fit for this?
Chat citations are flat numbers — readers cannot verify which sentence is grounded where.
Related patterns
- Often paired with: Confidence Signals — Sources per claim + uncertainty per claim — paired evidence and calibration.
- Often paired with: Reasoning Trace — Reasoning chain anchors at sources, not in synthesis prose.
- Alternative to: Smart Document — Sentence-level source mapping vs block-level summary blocks tied to a section.