# GDR-002 — C01 Version Transition (v1.2 → v1.5) **Governance Decision Record** **Programme:** PTODA Global Digital Authority Benchmark Series — C01 AI Crawler Access Study **Date:** 2026-06-29 **Author:** Douglas Lord · Digital Dominator Pty Ltd **Status:** FROZEN — decided and in effect, 2026-06-29. --- ## 1. Decision The C01 AI Crawler Access Study transitions from **v1.2** (the methodology and findings published on ptoda.org as of site version 1.3, dated 2026-06-06/2026-06-17) to **v1.5**, the four-market series frozen and published 2026-06-29. v1.5 **supersedes** v1.2 as the current research. v1.2 is **not deleted**: it is retained on the live site as part of the published historical record, each page carrying a clear superseded notice that links to its v1.5 successor. This is a supersession with preservation, not a silent overwrite. ## 2. Why v1.2 was superseded v1.2 reported AI-crawler restriction on a **binary basis**: a domain either blocks at least one AI retrieval crawler or it does not. On that basis it reported a pooled rate of ~40.2% across four markets (range ~33–42%) and led with the conclusion that "AI-crawler blocking is common in every market." Subsequent methodological work demonstrated that the binary measure is an **over-reading** of the underlying robots.txt policies. A binary count records any disallowed path as "blocking" — including operational and secondary paths (carts, search endpoints, APIs, administrative routes, asset bundles) that carry no substantive public content. Counting these as "blocking AI" conflates a site restricting its checkout API with a site withholding its primary public content. v1.5 replaces the binary with a **five-class functional-purpose taxonomy** that classifies each restricted path by what it is, and reports restriction at three thresholds: strict (whole-site exclusion), meaningful (primary content or whole-site), and expanded (any restriction including secondary/operational paths). It also separates declared robots.txt policy from infrastructure-layer access controls, and reports policy-observed denominators throughout. ## 3. What changed | Dimension | v1.2 | v1.5 | |---|---|---| | Restriction measure | Binary (block / no block) | Five-class functional taxonomy; strict / meaningful / expanded | | Headline framing | "Blocking is common" (~40.2% pooled) | "Binary overstates meaningful restriction" | | Whole-site exclusion (reported) | Not separated from binary | Strict: 3.8%–5.0% across markets | | Primary-content restriction | Not separated | Meaningful: 8.0%–11.7% across markets | | Markets | Pooled four-market binary | Four independently frozen country volumes + synthesis flagship | | Australia | Included in pooled count | Independent frozen volume (dataset `c01_au_409_v15_wired`, wired-confirmed) | | Instrument | C01 Crawler v1.2 | C01 Crawler v1.5 / v1.5.1 | | Architecture | Pooled analysis | Descriptive country volumes; inference reserved to the flagship | ## 4. What did NOT change - The underlying robots.txt observations are not in dispute; v1.5 re-reads the same kind of data with a finer instrument. The binary rates v1.2 reported remain arithmetically correct as *binary* rates — they are simply a coarser, and in the programme's current view misleading, way to summarise the policies. - The separation of declared policy from infrastructure-layer access (the "two-layer" model) originated in the v1.2-era work and is carried forward, refined, into v1.5. - The programme's author, publisher, methodology stewardship, and IP ownership are unchanged. ## 5. Why the published conclusions differ Because the *measure* changed, not because the data changed. v1.2's "blocking is common" (~40%) and v1.5's "binary overstates meaningful restriction" (whole-site 4–5%, primary-content 8–12%) are both true statements about the same domains, read at different thresholds. v1.5's position is that the binary headline, while not arithmetically false, materially overstates the degree to which businesses withhold their **primary public content** from AI systems — which is the question the public conversation actually cares about. v1.5 is therefore the programme's current and recommended reading. ## 6. Why both versions remain available Erasing v1.2 would be a silent rewrite of the published record. The programme's standing discipline is that superseded work is governed by formal rule and preserved, not deleted (Editorial Constitution §§21–22). Retaining v1.2 with a superseded notice: - preserves the provenance and publication history, - makes the methodological evolution legible to any reader, - explains *why* the headline changed rather than presenting a changed headline without account. This is the programme demonstrating that it updates its conclusions when better methodology becomes available, rather than defending earlier conclusions for the sake of consistency. ## 7. Implementation - **v1.5 pages become the canonical C01 research pages** at the canonical URLs: US, AU (new), GB, SG, and the cross-market flagship. - **v1.2 pages retained as archived historical records** with superseded notices linking to their v1.5 successors (at `-v12` URLs, noindex). - **C01 Version History page** created, listing v1.2 (superseded) and v1.5 (current). - **Homepage and research index** re-led with the v1.5 conclusion; the binary "2 in 5" headline moves into the historical record. - **Site version** increments to reflect the transition; the v1.2 access pages are no longer described as the current frozen access studies but as the historical version. - This record (GDR-002) is the authoritative explanation, linked from the version-history page and the superseded banners. ## 8. Relationship to prior records - **GDR-001** governed the v1.2→v1.5 *instrument/dataset* transition (the CSV read-limit fix and the v1.4→v1.5 classification work) at the data layer. GDR-002 governs the *publication* layer: how the live site moves from presenting v1.2 to presenting v1.5. - The four v1.5 country volumes and the flagship are frozen under the C01 Freeze Record (2026-06-29); their SHAs are recorded there.