Revenue System Audit / discovery layer
Can AI systems accurately understand your company?
Tony tests whether ChatGPT, Gemini, Google AI and other discovery systems can connect the company's identity, services, locations and credible evidence without guessing.
Ten evidence areas
Inspect sources—not a vanity score.
Each observation carries a date, source, confidence and limitation. The output is a prioritized implementation backlog, not a promise that an answer system will cite or rank the company.
| Audit area | Question tested | Evidence reviewed |
|---|---|---|
| Brand and entity consistency | Is the same organization clear across first- and third-party sources? | Names, descriptions, contact facts, profiles and organization relationships. |
| Schema and structured data | Do machine-readable facts match the visible page? | Person, Organization, Service, LocalBusiness, Article, FAQ and breadcrumbs where appropriate. |
| Service and location architecture | Can systems connect real offers to genuinely served places? | Canonical service/location pages, navigation, internal links and duplication. |
| FAQ coverage | Are important buyer questions answered directly and supportably? | Visible answers, factual consistency, context and linked evidence. |
| Third-party authority | Which independent sources corroborate identity, expertise or work? | Relevant profiles, mentions, reviews, directories and publications. |
| AI-readable content | Can systems extract company, offer, audience, geography and proof? | Headings, summaries, authorship, dates, crawlability and source links. |
| llms.txt | Does the optional publisher summary point to accurate sources? | Scope, links, freshness and consistency; no control or ranking claim is attached. |
| Citation and source-worthiness | Are material claims specific, useful and supportable? | First-party records, inspectable proof, external support and claim boundaries. |
| Competitor visibility | Which alternatives appear for the same need and why? | Dated answer patterns, cited domains, coverage and entity clarity. |
| Realistic customer prompts | What happens when likely buyers ask for this kind of provider? | Recorded wording, date, result, citations, geography and limitations. |
Problem → system → evidence
A decision-ready visibility diagnostic.
- Observed stateWhat each system and source can find on the recorded date.
- Identity mapCanonical names, services, locations, relationships and conflicting facts.
- Evidence backlogPages, schema, authority, FAQs and source links that need correction.
- Retest planReal prompts, crawl checks and dated observations after release.
Status: Awaiting baseline. No ranking, AI citation, traffic, lead or revenue result is claimed. Paid placement is evaluated separately and is never described as influencing organic recommendations.
Testing control contract
Record what was observed. Do not turn inference into fact.
- Purpose
- Evaluate entity clarity, source-worthiness and recommendation context using documented public evidence.
- Inputs
- Approved public pages, public profiles, schema, crawler responses, named competitors and realistic prompts.
- Allowed data
- Public business information and client-approved non-sensitive facts needed for the audit.
- Prohibited data
- Credentials, private analytics exports, customer records, unpublished strategy or personal data without written scope.
- Read permissions
- Public sources by default; private tools only when separately authorized with least privilege.
- Write permissions
- No publishing, profile edits, Search Console requests or schema changes occur without approval.
- Approval points
- The business approves canonical facts, public claims, corrections and every production release.
- Logs
- Prompt, date, provider, result, visible citations, source URL, observation and limitation.
- Failure behavior
- Mark blocked, unavailable, inconsistent or uncertain observations; never fill evidence gaps with generated claims.
- Human owner
- A named business owner confirms identity, services, locations and support for material claims.
- Disable control
- Prompt testing or private-source access can stop without affecting the public website.
Practical starting point
Bring the questions a buyer actually asks.
Share the canonical company name, real service area, priority offers, public sources and realistic recommendation prompts. Tony will define the evidence path and the smallest safe correction release.