Audit methodology

Evidence collection

We retrieve the public homepage through an SSRF-protected crawler, record metadata and structured data, and build a fixed versioned set of buyer-intent prompts from the site language and industry signals.

How assistants are queried

Each configured engine is asked the same buyer question a person would type into a chat. We do not ask the model to score itself. ChatGPT, Gemini and Perplexity are queried through OpenRouter with web grounding where the model allows it. That is an API approximation of those products, not a logged-in session on chatgpt.com. Google AI Overviews is measured only when a search evidence provider is configured; otherwise it is omitted, not simulated.

Score

The score is the mean of successful observations. A named mention is weighted by list position in the answer. A citation URL that points at the audited domain adds visibility even when the brand is not spoken. Unavailable engines are excluded from the mean and shown as not measured.

Limitations

• Model answers vary by time, locale, account and provider.
• Web search may be unavailable for a given model; those answers are model knowledge and are labeled as such.
• Citation pages are fetched when they are public HTTPS URLs. Failed fetches are left unverified.
• The score is a measurement snapshot, not a guarantee of future recommendations or revenue.

Corrections

Send evidence of an incorrect result to sturoxcompany@gmail.com with the report date, domain and affected query.