Citation Source Provenance Check (QA)

Citation Source Provenance Check (QA)

1. 01

   Pull the raw citation list behind every AI visibility score before it reaches a client deck

   A visibility percentage without the underlying cited domains is unauditable. Export the source URLs for each tracked prompt and confirm the platform is reporting domains, not just counts.

2. 02

   Verify that each cited domain actually resolves and still contains the claim attributed to it

   AI engines cite pages that later get edited, paywalled, or deleted. Spot-check ten citations per client and flag any that 404 or no longer support the quoted position.

3. 03

   Separate primary-source citations from explainer and aggregator citations

   Government, regulatory, and research domains now dominate ChatGPT citation rankings while explainer content has slipped, so a client ranking on paraphrase-heavy pages is exposed to a citation drop.

4. 04

   Cross-check the same prompt set across at least three engines and log where results diverge

   KIME, Pallix, and Ayzeo each cover a different engine roster, so a single-platform read can overstate or understate visibility. Record engine-by-engine variance rather than averaging it away.

5. 05

   Timestamp every data pull and note the platform's refresh interval

   Third-party monitoring carries latency risk. If a platform refreshes weekly, a client reacting to a competitor move on Tuesday may be reading Monday's snapshot.

6. 06

   Flag any metric sourced from a single vendor with no second confirmation

   Share-of-voice and sentiment scores are vendor-defined. Mark them as directional in reporting unless a second platform or a manual query confirms the direction.

7. 07

   Log corrections and reissue the affected client report within one business day

   A retracted visibility claim costs less trust than a quietly stale one. Note what changed, which engine or source drove it, and what the corrected figure is.