Cross-Engine Citation Reconciliation (QA)

Cross-Engine Citation Reconciliation (QA)

1. 01

   Freeze the query set before pulling any visibility numbers

   Lock 25 to 50 prompts that mirror the client's buying questions, then run the identical set across every engine in scope so week-over-week movement reflects the market rather than a changed prompt list.

2. 02

   Log the engine version and run date beside every visibility score

   A citation captured on one model build can vanish after the next release. Record the model name and timestamp so a client question about a sudden drop has an answer other than guesswork.

3. 03

   Separate brand mentions from citation sources

   A client can be named in an answer without being the source the engine cites. Track both columns, because the fix for a missing mention (content and schema) differs from the fix for a missing citation (third-party coverage).

4. 04

   Reconcile AI visibility against Search Console and analytics

   ChatGPT referral traffic frequently lands in Direct or Other buckets, so a flat organic line does not prove AI visibility is flat. Cross-check before reporting either number as the client's AI story.

5. 05

   Flag any metric that moved more than 20 percent in one cycle for manual review

   Large single-cycle swings usually trace to a prompt rewrite, an engine update, or a sampling change rather than a real ranking shift. Confirm the cause before it reaches a client deck.

6. 06

   Score each finding on evidence strength before it becomes a recommendation

   Label items as confirmed (seen across two or more engines or two or more runs), directional (single observation), or unverified. Only confirmed items earn a line in the retainer scope.

7. 07

   Archive the raw export with the report it produced

   Keep the underlying rows for at least two quarters. When a client disputes a trend six weeks later, the original data settles it faster than a re-run on a changed engine.