1. What a scan measures
A standard Cited scan generates 12 high-intent prompts from the submitted business type and market. The suite covers recommendations, urgency, cost, service selection, trust, and local comparison. The same prompt suite is submitted to ChatGPT Search, Gemini, and Google search queries eligible for AI Overviews through a third-party search-data provider.
We label those consumer surfaces explicitly. We do not substitute a developer model API and call it ChatGPT Search, Gemini's consumer product, or a Google AI Overview.
Every prompt names the submitted local market. Provider location controls differ by surface: Gemini and Google use the matching city, region, or country scope supported by the provider, while the ChatGPT consumer scraper currently uses country scope with the local market carried in the prompt. The requested market and provider scope are retained with the evidence.
2. Answer share
Answer share is the number of completed observations that mention the target business divided by the number of completed observations, multiplied by 100 and rounded to the nearest whole percent.
A mention is accepted when the normalized business name, the submitted domain, or a sufficiently distinctive normalized name variant appears in the captured answer. Cited records the reason for the match. It does not use fuzzy matching alone to turn an ambiguous phrase into a mention.
Failed requests, provider errors, and Google searches where no AI Overview appears are marked unavailable and excluded from the denominator. A report shows the numerator and denominator so that a percentage cannot hide a thin sample.
3. Competitors and citations
Named alternatives are extracted from structured provider fields when available and from the answer text otherwise. We remove the target business and obvious non-business labels. A competitor position is the order in which that name appears in the captured answer, not a permanent market rank.
Citations are the source links returned with the observation. Their presence does not prove that a source caused the answer, and their absence does not prove that the system used no outside source. Cited uses recurring citation domains as a prioritization signal, not as an attribution claim.
4. Website and business-fact audit
Cited reads a small set of public pages from the submitted origin, follows only public redirects, rejects private or local network targets, respects response-size limits, and inspects the site's robots file. It extracts visible names, services, locations, phones, hours, headings, and JSON-LD. Public business-profile data may be used to compare basic facts.
The initial report checks seven practical areas:
- recognized and internally consistent LocalBusiness structured data;
- direct, answerable service content;
- visible and consistent hours or availability;
- source domains recurring in captured answers;
- specific service-market coverage;
- name, address, phone, URL, and profile consistency where observable;
- support for any emergency or round-the-clock claims.
5. Drafted fixes
Cited generates fixes from extracted facts. Missing or conflicting details remain explicit placeholders or are marked for confirmation. Structured data should mirror visible page content and must be reviewed before publishing. Drafts are not legal, safety, licensing, pricing, or technical advice.
6. Why results change
Generated answers are stochastic and can change with wording, time, location, model updates, personalization, provider behavior, and the live web. One scan is a timestamped sample, not search volume, traffic share, a stable rank, or a forecast. Repeated scans can reveal directional movement, but do not establish causation.
7. What Cited does not promise
Cited does not promise inclusion, ranking, leads, revenue, or a specific percentage improvement. Neither valid schema nor any other single change guarantees inclusion in generative features. Search and AI providers control their own systems and can change them without notice.
8. Corrections and questions
If a report matched the wrong business or shows a reproducible data issue, email morais.aaron@gmail.com with the report link and the disputed observation.