AI Citation Coverage Audit (QA)

AI Citation Coverage Audit (QA)

1. 01

   Freeze the prompt set before any tool is connected

   Write 20 to 40 buyer-intent prompts per client, split across category, comparison, and problem-aware phrasing, and lock the list for the quarter so month-over-month movement reflects visibility rather than prompt edits.

2. 02

   Pull the same prompt set through at least two AI surfaces

   Run each prompt against ChatGPT and Perplexity as the minimum pair, then add Gemini or Google AI Overviews where the client's category skews to those surfaces. Nightwatch and LLMrefs both support multi-surface citation monitoring, so the audit can run from one dashboard rather than manual copy-paste.

3. 03

   Record mention rate, position, and sentiment as three separate numbers

   A brand can be mentioned often but ranked low inside the answer, or ranked first with negative framing. Peec and MaxAEO both expose these as distinct fields, which prevents a single blended score from hiding a sentiment problem in a client report.

4. 04

   Trace which external sources the AI answer cites

   Log the domains and pages the model leans on for each prompt. AIclicks and CiteRank surface citation sources directly, and that list becomes the outreach and content target list for the next delivery cycle.

5. 05

   Reconcile AI citation data against classic SERP rank data

   Compare the AI visibility baseline to positions pulled from AccuRanker or Nozzle for the same keyword themes. Divergence between page-one rankings and AI citation absence is the finding that justifies new scope, not a reporting error.

6. 06

   Flag volatility before it reaches the client deck

   Mark any prompt where the brand appeared or disappeared across two consecutive weekly pulls. AI answers shift with model updates, so a single-week swing is a monitoring note, not a strategy change.

7. 07

   Publish the baseline as a fixed-format appendix in the retainer report

   Keep the same prompt list, surfaces, and metric definitions every month so the client sees a trend line rather than a fresh narrative. TrackRight and Nightwatch both support white-label output if the agency resells the report under its own brand.