Appendix: Claim Basis Table · IDAEO: Marketing in the Age of AI

Appendix: Claim Basis Table

The claims in each chapter’s “Key Points of This Chapter” total 57, and each one is marked with what it rests on: First-Party Data (IDAEO’s own measurement records), Platform or Standards Documents (official documents from AI platforms and standards bodies), Academic Literature (the theoretical foundations and the numbered entries in the reference list), Rule of Thumb (industry or author rules of thumb, not measured on this site), and Theory Definition (IDAEO’s own definitions and forecasts). One claim can rest on several types at once. The classification was made independently by three reviewers, arbitrated against the data ledger, and then confirmed by the author. These are the basis types of the claims, which is a different matter from the Official Source, First-Party Source, and Third-Party Source of “Triangulated Verification” in Chapter 4.

Data Notes

• First-Party Data: taken from IDAEO's first-party measurement records. The period is Mar 3, 2026–Oct 8, old site and new site combined. AI, search engines, and Automated Bots total 105.5 million requests (105,487,618). This is consistent with IDAEO's approved basis. These are platform-site totals and cannot be compared directly with a single business's official website. Records before Aug 24 may include duplicate requests. AI Crawl Volume is 97.46 million, listed by the purpose of the requests. Copied back for training: 52.2 million. This means requests made for training. Building the AI index: 3.99 million (including 2.82 million from our own Cloudflare AI Search, to be deducted in next month's update). Building the search index: 41.26 million. The indexing figures only indicate requests made for indexing. These requests do not mean the content has been used in training or included in an index. Applebot, PetalBot, YouBot, and Kagibot are classified as Search Engine Crawlers. The classification follows each company's official documentation (checked Oct 14, 2026).
• AI Agents: 17.95 million is automated browsing from Aug 26, 2026–Oct 8, listed separately by the function of programs operating a browser; named AI Agents are 144,000, for a total of 18.09 million. Hermes, OpenClaw, Devin, Grok, and Muse are listed as still being tracked, and named-visit results will be published later; Manus has 14 visits.
• Humans: Organic Traffic of 310,000 is the number of click-throughs to the website from search or AI (mainly from Aug 26, 2026 onward), of which Google accounts for 268,000; Local Search is still being tracked and will be published later.
• Automated Bots provide Automated Information Services: “AI tools” refers to requests labeled AI by name and classified by source under other automated services. SEO tool crawlers (Semrush, Majestic, Moz, Ahrefs, etc.) and web page previews (Facebook, LinkedIn, X, etc.) are listed by platform-verified identity (Aug 24, 2026–Oct 8), counted separately, and fall outside the totals above. The identity and function of the 150,000 requests marked “Tracking in Progress” are being confirmed by IDAEO on an ongoing basis, and classification results will be published later.
• Search result impressions: the breakdowns in each figure are the search platforms' original values (Google 7.41 million, Bing 910,000, totaling 8.32 million); the overview's approximately 8.22 million follows IDAEO's approved basis, a difference of about 100,000. Indexed pages are the September 2026 value; Bing is the sum of its subfolders.
• TK Lin's personal IDA field test (Oct 3, 2026): 100 questions in five groups of 20 (identity, company, expertise, Washinmura, IDA). The five AI models were Perplexity Sonar, OpenAI GPT-5.4, Anthropic Claude Sonnet 5, Google Gemini 3.5 Flash, and xAI Grok 4.5, for 500 answers in total. 84% means this IDA appeared in the source list the AI retrieved (four of the five models used the same search interface; the questions named the author himself; a single case), corresponding to 419 answers. It appeared 439 times in the source lists, ranking first among all sources. For the 40 questions that name him directly (the identity and company groups), the figure was 91.5% (183 answers), with this IDA appearing 190 times. Oct 11 follow-up: of 16 Perplexity questions in four languages, 13 had this IDA among their sources. A single case cannot be generalized directly to businesses.
• Start-up time: about 30 days for a Warm Start is a rule of thumb; Figure 6-6, “Cold Start Timeline,” adds a measurement: crawlers from the same AI platform arrive in sequence, 30–35 days apart. About 90 days for a Cold Start is an industry rule of thumb; 14 days is the measured case of TK Lin's personal IDA.
• Field-test merchants: the first batch of field-test merchants and the second batch of field-test industries are all real operating businesses, and IDAEO has verified each one's identity; by agreement, this book lists only the industry, and merchant names stay in IDAEO's internal records.
• Publisher: IDAEO is operated by 株式会社和心 (Washin Co., Ltd.; corporate number 8011401020677; Minato-ku, Tokyo, Japan); author TK Lin.
• Theoretical foundations: all 17 existing sources have been verified one by one, with authors and years listed per the reference list; for six of them, verification covered the publisher's catalog entry and abstract. Goodhart was presented in 1975 and published in 1976. The theoretical basis is grouped into marketing fundamentals, generative engine and AI search research, AI customers and AI Agents, information retrieval and credibility, performance measurement, start-up and accumulation, official documentation of AI Platforms and agent protocols, and the theory behind the Seven Laws and the journey; see “Theoretical Foundations and References” at the end of the book for the full references.
• External AI Visibility metrics: compiled from vendor documentation and each company's public definitions verified on 2026-10-11; comparisons use the same measurement scope and metric definitions.
• All logos are trademarks of their respective owners and are used in this book to identify the AI Platforms.
• The nine Performance Metrics and the IDAEO Solutions use this book's naming; validation data are counted across all platforms, and single-merchant results are measured from that merchant's own data. The measurement methods for AI Mention Rate, AI Answer Accuracy Rate, and AI Return Visits are set; IDAEO continues to accumulate them and will publish them later.
• Same-domain terms within the Optimization Domains are grouped by this book's functional classification, following each term's context of use; AIO is linked to AEO, and both correspond to AI answers using content. AAO, AXO, AAIO, as well as GAIO and AISO, are listed per the AI Platforms' glossary pages verified on 2026-10-11; SXO, VSO, ASO, CRO, SMO, SGE, and LLM SEO follow common industry usage, and this round's item-by-item verification covered the AI Platforms' glossary terms above.
• The extended figures follow the visual framework of the introductory course “Website Traffic in the AI Era”; the numbers are compiled from the same first-party measurement records through Oct 8, 2026, and journey steps are always written as S1–S10. The statistical scope of this edition is defined by the items listed in each figure.
• Read each extended figure by its period: AI requests Mar 3–Oct 8; whole site Feb 12–Oct 8; automated browsing Aug 26–Oct 8; per-site breakdown Aug 20–Oct 8; Copilot citations Jul 22–Oct 8. Before Aug 24, first-party measurement records from multiple sources were merged and may include duplicate requests; when citing, note this scope as well.
• Copilot citations of 31,700 follow IDAEO's approved basis and cover the formal content scope; the raw aggregate is 52,600.
• “Reads for AI Answers,” 46.05 million, follows IDAEO's approved basis and includes search indexing and AI Overview generation, AI indexing, real-time retrieval, and named AI Agents. Of this, 41.26 million search-indexing reads are shared by search and AI answers; the rest are summed by function, presenting the scope of content use as a whole.
• Figure 6-6, “Cold Start Timeline,” shows the record of AI Platforms arriving in sequence. For five of the six AI Platforms, the first crawler and the next were 30–35 days apart. This is the gap between different crawlers from the same AI platform appearing in sequence. Crawlers whose first day is Mar 3, 2026 were already present when measurement began. The listed intervals are calculated from the observation start and represent the lower bound of the actual interval. A Cold Start is a new website's AI traffic accumulating from 0. The industry rule of thumb is about 90 days. A Warm Start has content that meets S1 and goes live on an AI Traffic Hub. This means a platform accumulating ten million or more AI requests per month. The IDAEO platform had 25.97 million in July, 35.99 million in August, and 15.79 million in September. Records before Aug 24 may include duplicate requests, so the counting basis differs before and after that date. The figure for September was lower than for August; keep the counting basis in mind when comparing months. The rule of thumb for a Warm Start is about 30 days. TK Lin's personal IDA case reached S7 in 14 days: live on Sep 20, tested on Oct 3 with 100 questions and five AI models. 84% means this IDA appeared in the source list the AI retrieved (four of the five models used the same search interface; the questions named the author himself; a single case). See the data notes for “TK Lin's Personal IDA Field Test” for limitations. The findings cannot be generalized to businesses.
• AI Agent Visits of 18.09 million = automated browsing 17.95 million + named AI Agents 144,000. A further 5,102 AI Agent visits came through ordinary browsers; these are counted separately and are not included in any total.
• Early test merchants, the second batch of test industries, and our own Washinmura official site: merchant-page search clicks, AI-referred customers, AI reads of merchant pages, and merchant official-site opens are taken from IDAEO first-party measurement records (queried 2026-10-10); AI exposure of the second-batch test industries' IDA pages is taken from IDAEO first-party measurement records (queried 2026-10-11, period Jul 18–Oct 8; summed across the 10 platforms or across the three purposes, each totals 373,776); search clicks and Copilot citations for industry official sites are taken from authorized search platform data (queried 2026-10-11), totals only. The Washinmura official site's 109,140 Google search clicks (Mar 3–Oct 8) are taken from search platform data; its 95,333 AI and search engine reads and 11,333 AI real-time retrievals (Aug 20–Oct 8) are taken from IDAEO first-party measurement records. The Google comparison with the same month last year includes only official sites with last year's records; in the Bing total, data for some official sites starts from Aug 13.