Occupation manual 007 / Search & answer visibility
A proof-first route from one real query to a direct answer, source coverage, crawl/index controls, structured-data parity, internal links, submission receipts, observed citations, and evidence-based refreshes.
Short answer: answer one real customer question better than a thin summary. Put the direct answer and useful artifact in crawlable HTML, tie material claims to visible sources, keep canonical/sitemap/robots/schema signals consistent, link the page into a clear site hierarchy, notify approved discovery systems, then measure observed indexing, referrals, and citations without promising any of them.
Bing's current webmaster guidance says the same crawl, indexing, URL consolidation, content clarity, and trust foundations used for search also support eligibility for Copilot grounding and citations. Google likewise emphasizes helpful, reliable, people-first content and clear technical access.
OpenAI documents OAI-SearchBot for ChatGPT search and GPTBot for potential model training as separate controls. Anthropic similarly documents Claude-SearchBot, ClaudeBot, and Claude-User roles. The site owner—not a generic SEO tool—chooses the intended access policy.
Name the question in the audience's language, the task or decision behind it, evidence that it is real, the existing page that may already answer it, and the unique artifact or analysis worth adding.
Draft a direct answer, audience, outcome, numbered route, artifact, evidence desk, limitation, related guides, and visible author/reviewer/update date.
Link each exact visible claim to evidence, class, scope, date, contradiction or limitation, and reviewer status. Search visibility never substitutes for claim truth.
Record intended access and actual robots, meta/header, firewall/CDN, and verification results for Google, Bing, OpenAI search/training, Anthropic search/training, and user-triggered agents where applicable.
Check server status, HTTPS canonical, redirects, sitemap entry and honest lastmod, index controls, title, description, H1, visible answer, structured data, assets, and rendering.
Use crawlable, descriptive internal links inside a clear hierarchy. Consolidate duplicate intent instead of creating doorway, location, job, or query variants with little added value.
Bind the exact page version to local QA, deployment approval/result, sitemap and inspection actions, IndexNow receipt, crawler access checks, errors, and limits. Each action needs its own authority.
Log surface, query, observed result/citation/referral, exact page version, evidence, interpretation limits, and next action. Update, consolidate, redirect, noindex, or remove based on evidence.
| Job | Approved input | AI output | Human gate |
|---|---|---|---|
| Query-intent clerk | Real customer language and observed search data | Question/job map and duplicate-intent flags | Content owner chooses whether a page is needed |
| Answer-brief clerk | Approved sources and artifact | Direct answer, route, limitations, internal links | Qualified reviewer approves claims and usefulness |
| Technical audit clerk | Page HTML, headers, sitemap, robots, current docs | Parity and conflict report | Site owner chooses and implements controls |
| Structured-data clerk | Visible reviewed page content | Schema proposal and validation report | Human confirms exact visible-content parity |
| Measurement clerk | Search-console, webmaster, analytics, and citation observations | Visibility log and refresh suggestions | Owner interprets limits and approves changes |
OpenAI's current crawler page says OAI-SearchBot is used for ChatGPT search, while GPTBot covers content that may be used to improve foundation models. Anthropic documents Claude-SearchBot, ClaudeBot, and Claude-User separately. Treat each published crawler purpose as its own owner decision. Training access is not a prerequisite you should assume for search access.
Use a supported vocabulary and the search engine's current feature rules. Mark up only content that exists visibly and truthfully. Never generate fake authors, ratings, reviews, FAQs, prices, products, organizations, dates, or claims to create a richer machine-facing version than readers receive.
A sitemap helps discovery and should list canonical URLs with accurate freshness signals. Bing recommends IndexNow for notifying participating engines when URLs are added, changed, or deleted, but its own documentation says submission does not guarantee crawl or index. Record `submitted`, `received`, `crawled`, and `indexed` separately.
Google's helpful-content guidance asks whether a page adds original information, reporting, research, or analysis. Bing says content that is discoverable, accurate, well-structured, and created for users performs best across search experiences. For StackPilot, that means one direct answer, one practical route, one real artifact, named sources, visible limitations, and related guides—not thousands of generated occupation shells.
Act as a search and answer visibility preparation clerk, not a search engine, crawler owner, publisher, deployer, ranking authority, or citation guarantor. Using the supplied page, sources, and current first-party documentation, create: (1) a real-query/job card, (2) an answer-first content brief, (3) a claim/source coverage table, (4) a crawl/index control matrix separated by crawler purpose, (5) canonical/sitemap/structured-data parity checks, (6) an internal-link/entity map, and (7) a measurement and refresh plan. Do not invent search volume, rank, traffic, citations, reviews, authorship, dates, schema fields, or crawler behavior. Do not publish, deploy, submit URLs, change robots, or modify search accounts. Mark every assumption and recheck requirement.
Download the desk, audit a real guide, and require evidence at every state: useful, published, crawlable, indexed, surfaced, cited, referred, and completed.