AI Answer Baseline Audit (Onboarding)

AI Answer Baseline Audit (Onboarding)

1. 01

   Freeze the client's current AI answer set before any optimization work begins

   Run 20 to 30 buyer-intent prompts through ChatGPT, Perplexity, Gemini, and Google AI Overviews, then export the raw responses with timestamps. This becomes the baseline that every later claim of improvement is measured against.

2. 02

   Log which sources the models actually cite for each prompt

   Record the domain, page URL, and whether the client appears as a primary citation, a passing mention, or not at all. Citation source analysis is the part most agencies skip, and it is the part that determines whether the fix is content or off-site.

3. 03

   Score sentiment per mention on a three-point scale

   Mark each appearance as accurate, neutral, or misrepresented. A single misrepresented mention in a high-intent prompt matters more than ten accurate mentions in low-intent ones.

4. 04

   Separate hallucination findings from ranking findings

   A model inventing a service the client does not offer is a different problem from a model omitting the client entirely. Route the first to a correction plan and the second to a visibility plan.

5. 05

   Map every finding to a named owner and a delivery date

   Assign content rewrites, review responses, and citation outreach to specific people. Findings without owners become a slide in the kickoff deck and nothing else.

6. 06

   Confirm the client's review profile is claimed and current on the platforms models read most

   Google Business Profile, Facebook, and industry directories feed AI answers indirectly. An unclaimed profile is a silent gap in the baseline.

7. 07

   Set the re-audit date at 30 days and state the success threshold in writing

   Agree on a target such as reducing misrepresented mentions from four to one, or appearing as a primary citation in three of the tracked prompts. Vague improvement language invites retainer disputes.