{
  "_correction_2026_09_05": {
    "by": "bg-d-bq-1888-cd0458, 2026-09-05, after GPT Pro review pr-20260905T055221Z-282def finding 14",
    "supersedes": [
      "any line in this file's commentary saying no language model crawler (training or search) appears in any window",
      "any line inferring that the Gemini Notebook fetcher renders pages or executes JavaScript"
    ],
    "text": "Training crawlers ARE present: this file's own classifier_families record Meta-ExternalAgent 49 and Bytespider 37 rows (bin ai_crawler, 86 in all). The statement that holds is narrower: no crawler from OpenAI, Anthropic, Perplexity, Cohere or Common Crawl appears in any window. The three Google-NotebookLM rows show that requests under that token reached the beacon's endpoint (POST /api/views); they do not show that the fetcher rendered the page or executed its script, since a direct call to the endpoint leaves the same row. No measured count in this file was altered."
  },
  "_note": "Durable aggregate snapshot of the LIVE page_views window, taken 2026-08-30 by bg-c-bq-072-96d241 (burn-guard, account C, read-only). No IP addresses and no raw user-agent strings are stored (privacy rule). Third file in the series after bg-d-bq-072/evidence/page-views-export-2026-07-04-aggregates.json (window A, Feb 11 - Jun 9) and bg-d-bq-071-work/page-views-live-2026-08-20-aggregates.json (window B, May 24 - Aug 19). Rows here overlap window B from 2026-06-01 onward; production has purged May 24 - May 31 since. Classification is the shipped classifier (counter-patch/api/src/lib/audience.ts, 2026-08-30 revision), run over the distinct user_agent groups of the whole table, not a hand grouping.",
  "captured_at": "2026-08-30",
  "captured_at_note": "taken between 10:20Z and 10:29Z on 2026-08-30; the exact minute was not recorded (the original field carried the placeholder '2026-08-30T10:2xZ', replaced 2026-09-05)",
  "source": "vault exec --as aox-production-repair -- wrangler d1 execute ashitaorbis-db --remote --json (three read-only SELECTs; every meta returned rows_written 0, changed_db false)",
  "window": {
    "oldest_row": "2026-06-01T15:16:35.281Z",
    "newest_row": "2026-08-30T03:44:57.250Z",
    "rows_total": 714,
    "distinct_paths": 66
  },
  "as_flagged_by_production": {
    "is_ai_1": 489,
    "is_ai_0": 225,
    "flagged_ai_share_pct": 68.5
  },
  "classifier_bins": {
    "search_crawler": 403,
    "other_bot": 96,
    "ai_crawler": 86,
    "human": 126,
    "ai_agent": 3
  },
  "classifier_families": {
    "Googlebot": 286,
    "browser": 126,
    "Applebot": 96,
    "GoogleOther": 72,
    "Meta-ExternalAgent": 49,
    "Bytespider": 37,
    "headless-browser": 21,
    "Bingbot": 21,
    "Gemini-Notebook": 3,
    "AhrefsBot": 2,
    "QuillBot": 1
  },
  "machine_share_pct": 82.4,
  "production_flag_vs_classifier": {
    "agree": 615,
    "disagree": 99,
    "detail": {
      "GoogleOther: prod is_ai=0 -> other_bot": 72,
      "headless-browser: prod is_ai=0 -> other_bot": 21,
      "Gemini-Notebook: prod is_ai=0 -> ai_agent": 3,
      "AhrefsBot: prod is_ai=0 -> other_bot": 2,
      "QuillBot: prod is_ai=0 -> other_bot": 1
    }
  },
  "delta_since_2026_08_20_snapshot": {
    "_comment": "rows with created_at > 2026-08-19T23:51:57.458Z (the 08-20 snapshot's newest row)",
    "rows": 172,
    "bins": {
      "search_crawler": 85,
      "ai_crawler": 33,
      "human": 46,
      "other_bot": 6,
      "ai_agent": 2
    },
    "families": {
      "Googlebot": 61,
      "browser": 46,
      "Bytespider": 33,
      "Applebot": 19,
      "Bingbot": 5,
      "headless-browser": 3,
      "Gemini-Notebook": 2,
      "GoogleOther": 2,
      "AhrefsBot": 1
    },
    "machine_share_pct": 73.3
  },
  "language_model_fetchers": {
    "Gemini-Notebook (Google-NotebookLM token)": {
      "rows_total": 3,
      "rows_inside_window_B_2026_05_24_to_08_19": 1,
      "first_row": "2026-07-29T15:59:10Z",
      "rows_after_08_20_snapshot": 2,
      "last_row": "2026-08-23T12:04:46Z",
      "stored_is_ai": 0,
      "paths": [
        "012-red-teaming-your-business-ideas",
        "009-ai-evaluating-ai"
      ],
      "_comment": "Bare token, no bot-shaped substring, stored as human by production AND filed under 'plausible_browser' by the 08-20 hand derivation. Google documents Google-NotebookLM as the former token (supported until August 2026) of the Gemini Notebook user-triggered fetcher, which requests URLs a user handed the product as a source."
    },
    "GPTBot": 0,
    "ClaudeBot": 0,
    "Claude-Web": 0,
    "anthropic-ai": 0,
    "ChatGPT-User": 0,
    "Claude-User": 0,
    "PerplexityBot": 0,
    "Perplexity-User": 0,
    "CCBot": 0,
    "cohere-ai": 0,
    "OAI-SearchBot": 0,
    "Google-GeminiNotebook (new token)": 0,
    "Google-Agent": 0
  },
  "correction_to_window_B_file": {
    "plausible_human": "81 -> 80",
    "llm_assistant_agent_crawlers": "0 -> 1 (Gemini-Notebook, 2026-07-29)",
    "machine_share_pct": "85.4 -> 85.6",
    "_comment": "The 08-20 file is not edited (archive-don't-overwrite; another session's report path). This record is the correction."
  },
  "replication_vs_prior_windows": {
    "feb11_jun09_export": {
      "machine_share_pct": 79.1,
      "llm_fetcher_rows": 0
    },
    "may24_aug19_live_corrected": {
      "machine_share_pct": 85.6,
      "llm_fetcher_rows": 1
    },
    "jun01_aug30_live_this_file": {
      "machine_share_pct": 82.4,
      "llm_fetcher_rows": 3
    },
    "verdict": "Machine share high in all three reads; no language model CRAWLER (training or search) in any; the only model-side traffic the beacon has ever recorded is Google's notebook fetcher, three rows, which renders because it runs on Google's crawl infrastructure and which the pattern list filed as human."
  }
}