{
  "metric": "Retrieval Token Cost (RTC)",
  "page": "/methodology/retrieval-token-cost/2026-04-27",
  "frozen_date": "2026-04-27",
  "publisher": "GEOlocus.ai (GeoLocus Group, a subsidiary of Aryah.ai)",
  "formula": "RTC = (response_tokens * ttlb_seconds) / useful_chars",
  "perspective_caveat": "Absolute TTLB / RTC values are from a residential Phoenix AZ measurement (10 rapid-fire hits per host, pinned CF edge IP, compressed). RATIOS (Top10Lists.us vs cohort) are network-locale invariant. Final published absolute numbers will refresh when CF Worker datacenter measurement ships.",
  "anonymization_note": "Cohort sites are anonymized as Site A through Site D in the rendered methodology page. This receipts.json names them. Top10Lists.us is named on the page as the flagship + proof-of-concept.",
  "tokenizer": "tiktoken cl100k_base (OpenAI / GPT base encoding); chars/4 approximation used where exact tiktoken count unavailable",
  "useful_chars_source": "Same MAX(boilerplate-strip, trafilatura) extractor as the Relevance Ratio benchmark; primary_content_chars from the 2026-04-27 RR per-page table (homepage row)",
  "ttlb_source": "C:/Users/ROBER/Downloads/seo-cohort-perf-2026-04-27.json (10-hit rapid-fire pinned-IP residential Phoenix benchmark)",
  "cohort": [
    {
      "label": "Top10Lists.us",
      "domain": "www.top10lists.us",
      "role": "flagship + proof-of-concept",
      "ip_pinned": "172.66.174.215",
      "response_tokens_estimated": 2596,
      "ttlb_p50_seconds": 0.115,
      "ttlb_mean_seconds": 0.118,
      "useful_chars_homepage_post_fork": 5954,
      "rtc": 0.05,
      "ratio_vs_top10": 1.0,
      "notes": "Bot-facing edge function returns clean-room HTML: no React shell, no analytics tags, no third-party trackers, no font-loader scripts"
    },
    {
      "label": "Site A",
      "domain": "moz.com",
      "ip_pinned": "104.16.127.24",
      "response_tokens_estimated": 31250,
      "ttlb_p50_seconds": 0.123,
      "ttlb_mean_seconds": 0.123,
      "useful_chars_homepage": 11853,
      "rtc": 0.32,
      "ratio_vs_top10": 6.4
    },
    {
      "label": "Site B",
      "domain": "seo.com",
      "ip_pinned": "104.18.28.239",
      "response_tokens_estimated": 52830,
      "ttlb_p50_seconds": 0.204,
      "ttlb_mean_seconds": 0.208,
      "useful_chars_homepage": 12123,
      "rtc": 0.89,
      "ratio_vs_top10": 17.8
    },
    {
      "label": "Site C",
      "domain": "brafton.com",
      "outcome": "blocked",
      "outcome_evidence": "HTTP 403 to all bot UAs; this RTC is browser-UA best case; AI-bot-observed RTC is N/A",
      "response_tokens_estimated": 29420,
      "ttlb_p50_seconds_browser_ua_estimated": 0.180,
      "useful_chars_homepage_browser_ua": 11268,
      "rtc_browser_ua_best_case": 0.47,
      "ratio_vs_top10": 9.4
    },
    {
      "label": "Site D",
      "domain": "victorious.com",
      "ip_pinned": "162.159.134.42",
      "response_tokens_estimated": 47180,
      "ttlb_p50_seconds": 0.195,
      "ttlb_mean_seconds": 0.206,
      "useful_chars_homepage": 21047,
      "rtc": 0.44,
      "ratio_vs_top10": 8.8
    }
  ],
  "limitations": [
    "Residential measurement; CF Worker datacenter measurement in flight",
    "Tokenizer choice: cl100k_base; per-engine RTC would shift modestly with different encodings",
    "Homepage only; deep-content RTC behaves similarly but absolute numbers differ",
    "RTC depends on RR; if RR regresses, RTC rises",
    "Site C bot-block: browser-UA best case; AI-bot-observed RTC is N/A"
  ]
}