Methodology · Joe Starcher

How We Evaluate AI Visibility

An AI visibility audit should explain what was reviewed, how observations were made, and which conclusions cannot be supported. It should not turn a single prompt into a permanent “ranking.”

1. Establish the business entity

We begin with the approved business name, website, ownership or leadership, contact information, services, service areas, and important profiles. The goal is to identify whether the same organization is described consistently across its own pages and relevant public sources.

2. Review crawl and index foundations

We review public accessibility, robots directives, canonical signals, sitemap coverage, page titles, headings, internal links, and structured data. For ChatGPT Search discoverability, OpenAI advises publishers not to block OAI-SearchBot. Google states that ordinary SEO fundamentals remain relevant to its AI features.

3. Evaluate service and location clarity

We look for clear pages explaining what the business does, who each service is for, which markets it serves, what buyers should expect, and which claims require qualification. A system cannot confidently interpret information that is vague, contradictory, or buried.

4. Evaluate answer-ready content

We compare website content with real questions buyers ask during discovery and evaluation. Strong content gives a direct answer, explains context, states limitations, and connects the answer to verifiable business expertise. Repeated keywords without useful substance do not satisfy this standard.

5. Review authority and corroboration

We identify relevant profiles, citations, reviews, interviews, organizational memberships, expert articles, and other legitimate sources. The audit does not treat every directory as valuable and does not recommend fabricated profiles, purchased reviews, or fake mentions.

6. Use a documented prompt set

When sampling AI answers, we record the platform, date, location context when relevant, exact prompt, and whether web search was used. We test several question types rather than one preferred phrase. Results can vary by platform, time, user context, and available sources.

7. Separate observation from attribution

An appearance in an AI answer is an observation. It is not automatic proof that one page, schema change, or service caused the result. Traffic and sales require separate analytics and attribution. OpenAI notes that ChatGPT referral links can include utm_source=chatgpt.com, which can help identify visits from ChatGPT search.

What the audit does not promise

  • No guaranteed mention, citation, ranking, traffic, lead, sale, revenue, or ROI.
  • No claim of direct submission to or control over an AI platform.
  • No permanent score based on one prompt or one day.
  • No fake scientific certainty where platform behavior is variable.

Primary references