Canonical
The clean commercial question used as the family baseline.
The ProjectSEO
Services
Industries
Locations
Research method · Updated 29 July 2026
Measure AI visibility without inventing an AI ranking.
Use a controlled prompt panel, repeat every condition and record the answer, sources, competitors, accuracy, referrals and qualified demand. A mention is an observation. It becomes useful only when the context and outcome stay attached.
TheProjectSEO registry
50 → 200
Fifty commercial prompt families, each expanded into canonical, persona, evidence and decision phrasings.
Direct answer
Prompt variation is not noise to remove. It is a condition to control.
ChatGPT can rewrite one question into several targeted searches. Google describes a similar query fan-out process for its AI features. Claude may use live web search and location context. That means wording, model, interface, market and time can change the sources and shortlist. We preserve those conditions and run the same prompt three times before comparing it with another period.
Persona
The same need with buyer role and company context added.
Evidence
The same need with explicit proof and validation criteria.
Decision
The same need framed as a comparison or next-step decision.
Observation record
Keep the answer and its operating conditions together.
Aggregate visibility scores hide the reason an answer changed. These fields keep the raw observation available for audit, diagnosis and commercial attribution.
01Exact prompt and prompt-family ID
02Platform, model and user interface
03Market, language and location context
04Whether live web search was active
05Timestamp and repeat number
06Brand inclusion and shortlist order
07Recommendation versus passing mention
08Cited brand URL and every source URL
09Visible search queries or query fan-out
10Competitors included in the same answer
11Factual accuracy and sentiment
12AI referral sessions, leads and revenue
What can improve source selection
Build pages worth retrieving, citing and trusting.
Google says there are no additional technical requirements or special AI schema files for its AI features beyond normal Search eligibility. OpenAI, Anthropic and Perplexity document search crawlers that publishers can allow. The practical work remains strong search fundamentals plus sources that add first-hand information a generated answer can verify.
Answer the real decision
Give the buyer a direct answer, comparison criteria, limitations and a useful next step. Do not manufacture separate thin pages for every prompt wording.
Publish original evidence
Use named methods, dates, screenshots, definitions, before-and-after validation and project-specific caveats. Commodity summaries give a retrieval system little reason to choose you.
Make entities unambiguous
Keep the company, people, services, projects, locations and evidence consistently named and internally connected. Correct inaccurate facts at the source rather than trying to manipulate an answer.
Protect crawl and index access
Verify Google and Bing indexability and make an explicit choice for OAI-SearchBot, Claude-SearchBot, PerplexityBot and their user-fetch agents. Training controls are not the same as search visibility controls.
What Expressway.PH teaches us
Task ownership can travel from search results into generated answers.
Expressway.PH was organized around real driver tasks—tolls, routes, exits, RFID systems, restrictions, traffic and policy updates—rather than a disconnected volume of generic articles. A supplied July 2026 Ahrefs snapshot showed 49.2K estimated monthly organic traffic, 6.8K keywords, 3.4K AI Overview responses and 2.4K ChatGPT responses.
Those figures are point-in-time third-party observations. They are not audited revenue, visitor counts from the AI platforms or a guaranteed outcome.
Review the Expressway.PH evidenceApplied to TheProjectSEO
- 01 · Own the buying task. Map every commercial prompt family to one credible page instead of creating a page for every wording.
- 02 · Show the work. Connect methodology, pricing, named project evidence, people and implementation scope.
- 03 · Observe the answer. Run three times, retain citations and competitors, and separate mention from recommendation.
- 04 · Measure demand. Join GSC, Bing, server logs, AI referrals, qualified forms, pipeline and revenue.
Primary sources
Platform claims should remain traceable.
Google Search: AI features and your websiteEligibility, query fan-out and Search Console measurementGoogle: Optimizing for generative AI featuresCurrent guidance on unique, useful and non-commodity contentOpenAI: ChatGPT searchQuery rewriting, source links and OAI-SearchBot accessOpenAI: Publishers and developers FAQChatGPT referral attribution and search crawler controlsAnthropic: Web search in ClaudeLive web search, citations and location-sensitive answersPerplexity: Crawler documentationSearch indexing, user fetches and crawler verification
Questions the method should answer
Can a company rank number one in ChatGPT?
Not in the same stable, query-position sense used for a conventional search result. Generated answers can change with the prompt, model, interface, market, web-search state and time. The defensible unit is a repeated observation under a recorded condition.
Why use 200 prompts instead of one brand query?
The 200-prompt registry begins with 50 commercial prompt families and expands each into canonical, persona, evidence and decision variants. That exposes wording sensitivity while keeping every variation tied to the same buying problem and page owner.
Should every prompt run on every platform and in every market?
No. That creates an expensive Cartesian product without guaranteeing a better decision. Use a small weekly sentinel panel, rotate the remaining variants across platforms and markets, and run event-based checks after material releases or model changes.
Does an AI citation prove commercial impact?
No. A citation is an observable source-selection signal. It should be joined with AI referrals, assisted conversions, qualified leads and revenue before it is treated as a business outcome.
Need the implementation?
Start with the prompt families tied to qualified demand.
We will map the questions, page owners, source gaps and measurement conditions before recommending content or technical changes. The goal is not more prompt screenshots. It is attributable search demand across Google and generated answers.
Review AI visibility tracking