# EASD 2026 study: post-registration robustness analyses (D24)

> **POST-REGISTRATION ROBUSTNESS (D24) — never replaces a registered verdict.** Every analysis below was added after the registered analysis ran. None replaces a registered figure or verdict; each answers a question the internal reviews of 11 Oct 2026 raised. The registered analysis is `scripts/easd-2026-study/analyze.ts` (unmodified) and its outputs `docs/evidence/easd-2026-analysis.md` and `docs/evidence/computed/easd-2026-analysis.json`.

Generated by `scripts/easd-2026-study/robustness.ts` from the frozen pull (sha256 `de98a60469b0e7b1…`, matched to `docs/evidence/computed/easd-2026-freeze.json`, commit `84a2214c`) and the frozen repository files, each matched to the freeze record by sha256. The database was read only for each admitted cycle's scan input (brand names, aliases, category, competitors), exactly as `analyze.ts` reads it. No scan, no paid API call and no database write.

**Validation before any figure.** The production scorer reproduced every governed per-cycle naming count (64 admitted cycles × 4 engines, 0 mismatches). The E1 new-URL and unreadable-page sets, rebuilt with the local Gemini resolution cache that the frozen E1 tally used, reproduced the frozen E1 inspection coverage and every per-cycle unverified count exactly (the cache is not in the freeze record; this reproduction is its check). Reading I1 reproduced every registered bridged band and verdict, the attribution counts reproduced every registered I3 test, and the Saudi E3-i shares reproduced the frozen exploratory figures.

## 1. Registered §5 descriptive source analyses — registered §5, reported late under D24

Protocol §5 registered three descriptive source analyses. The frozen analysis did not report them and no deviation recorded the omission (methodologist M5(f), M9(c)(ii); GEO reviewer point 3). They are reported here, late, as registered descriptives.

### 1(a) The Note 04 clinic page: Perplexity citations per cycle

Answers citing the page (exact normalised URL) over the engine's captured answers in each admitted UAE cycle. The page and its publisher are not named. Gemini's page-level match needs resolved redirect links, which exist for the congress and after windows only.

| Cycle date | Window | Perplexity citing / answers | Perplexity citing and naming | Claude | ChatGPT | Gemini |
|---|---|---|---|---|---|---|
| 2026-08-22 | before-v4 | **0/54** (0.0%) | 0 | 0/60 | 0/60 | n/a |
| 2026-08-24 | before-v4 | **0/60** (0.0%) | 0 | 0/60 | 0/60 | n/a |
| 2026-08-26 | before-v4 | **0/60** (0.0%) | 0 | 0/60 | 0/60 | n/a |
| 2026-08-28 | before-v4 | **0/60** (0.0%) | 0 | 0/60 | 0/60 | n/a |
| 2026-08-30 | before-v4 | **0/59** (0.0%) | 0 | 0/60 | 0/60 | n/a |
| 2026-09-04 | before-v4 | **8/60** (13.3%) | 8 | 0/60 | 0/60 | n/a |
| 2026-09-09 | before-v4 | **9/60** (15.0%) | 9 | 0/60 | 0/60 | n/a |
| 2026-09-11 | before-v4 | **8/60** (13.3%) | 8 | 0/60 | 0/60 | n/a |
| 2026-09-14 | before-v4 | **7/57** (12.3%) | 7 | 0/60 | 0/60 | n/a |
| 2026-09-16 | before-v4 | **12/56** (21.4%) | 11 | 0/60 | 0/60 | n/a |
| 2026-09-21 | before-v4 | **10/59** (17.0%) | 10 | 0/60 | 0/60 | n/a |
| 2026-09-23 | before-v4 | **7/60** (11.7%) | 7 | 0/60 | 0/60 | n/a |
| 2026-09-25 | before-v4 | **4/60** (6.7%) | 4 | 0/60 | 0/60 | n/a |
| 2026-09-27 | before-v5 | **3/58** (5.2%) | 3 | 0/60 | 0/60 | n/a |
| 2026-09-29 | during | **3/60** (5.0%) | 3 | 0/60 | 0/60 | 0/60 |
| 2026-10-04 | after | **4/60** (6.7%) | 4 | 0/60 | 0/60 | 0/60 |
| 2026-10-06 | after | **7/60** (11.7%) | 4 | 0/60 | 0/60 | 0/60 |
| 2026-10-07 | after | **3/60** (5.0%) | 3 | 0/60 | 0/60 | 0/60 |
| 2026-10-08 | after | **1/60** (1.7%) | 1 | 0/60 | 0/60 | 0/60 |
| 2026-10-09 | after | **2/60** (3.3%) | 2 | 0/60 | 0/60 | 0/60 |
| 2026-10-10 | after | **3/60** (5.0%) | 3 | 0/60 | 0/60 | 0/60 |
| 2026-10-11 | after | **4/60** (6.7%) | 4 | 0/60 | 0/60 | 0/60 |

**What this shows.** The page first appears in Perplexity's citations in the 2026-09-04 cycle, peaked at 12 answers in one cycle, and appears in 7 of 7 after-window cycles (4, 7, 3, 1, 2, 3, 4 answers). **What it does not show.** Whether the page caused any naming change: citation is a link, not an effect, and the counts are not compared with a band.

### 1(b) New local or regional pages drawing ≥ 10% of one engine's answers in a cycle

Congress- and after-window cycles. A page is NEW when no before-window cycle of the same series cited it (E1's manifest; Gemini's pre-congress redirects were never resolved, so Gemini over-includes, D16). LOCAL OR REGIONAL: the host is under a UAE, Saudi, Singapore or other GCC suffix, or is classified clinic or consumer/retail. A page is listed when the answers citing it are ≥ 10% of that engine's captured answers in the cycle. Pages are typed and given an opaque id, never named.

**ChatGPT, Claude and Perplexity** (every listed page-cycle):

| Series | Cycle date | Engine | Page (opaque id) | Type | Answers citing / answers | Share | Naming among citing |
|---|---|---|---|---|---|---|---|
| saudi-t1d | 2026-10-05 | ChatGPT | P17 | consumer / retail (incl. pharmacy) page, non-regional domain | 7/56 | 12.5% | 1/7 |
| singapore-sglt2 | 2026-10-02 | Perplexity | P24 | consumer / retail (incl. pharmacy) page, non-regional domain | 7/26 | 26.9% | 7/7 |
| singapore-sglt2 | 2026-10-02 | Perplexity | P25 | consumer / retail (incl. pharmacy) page, non-regional domain | 5/26 | 19.2% | 5/5 |
| singapore-sglt2 | 2026-10-02 | Perplexity | P26 | other (classified) page, Singapore domain (public: data.gov.sg) | 4/26 | 15.4% | 4/4 |
| singapore-sglt2 | 2026-10-02 | Perplexity | P27 | clinic / hospital / care provider page, non-regional domain | 4/26 | 15.4% | 4/4 |
| singapore-sglt2 | 2026-10-05 | Perplexity | P24 | consumer / retail (incl. pharmacy) page, non-regional domain | 7/26 | 26.9% | 7/7 |
| singapore-sglt2 | 2026-10-05 | Perplexity | P35 | clinic / hospital / care provider page, non-regional domain | 4/26 | 15.4% | 4/4 |
| singapore-sglt2 | 2026-10-05 | Perplexity | P36 | regulator / government health body page, Singapore domain (public: hsa.gov.sg) | 3/26 | 11.5% | 3/3 |
| singapore-sglt2 | 2026-10-05 | Perplexity | P37 | unclassified page, Singapore domain | 3/26 | 11.5% | 3/3 |
| singapore-sglt2 | 2026-10-06 | Perplexity | P24 | consumer / retail (incl. pharmacy) page, non-regional domain | 7/26 | 26.9% | 7/7 |
| singapore-sglt2 | 2026-10-06 | Perplexity | P35 | clinic / hospital / care provider page, non-regional domain | 4/26 | 15.4% | 4/4 |
| singapore-sglt2 | 2026-10-06 | ChatGPT | P42 | consumer / retail (incl. pharmacy) page, non-regional domain | 3/26 | 11.5% | 3/3 |
| singapore-sglt2 | 2026-10-06 | ChatGPT | P43 | consumer / retail (incl. pharmacy) page, non-regional domain | 3/26 | 11.5% | 3/3 |
| singapore-sglt2 | 2026-10-07 | ChatGPT | P45 | consumer / retail (incl. pharmacy) page, non-regional domain | 4/26 | 15.4% | 4/4 |
| singapore-sglt2 | 2026-10-09 | ChatGPT | P42 | consumer / retail (incl. pharmacy) page, non-regional domain | 3/26 | 11.5% | 3/3 |
| singapore-sglt2 | 2026-10-10 | ChatGPT | P45 | consumer / retail (incl. pharmacy) page, non-regional domain | 4/26 | 15.4% | 4/4 |
| singapore-sglt2 | 2026-10-10 | ChatGPT | P43 | consumer / retail (incl. pharmacy) page, non-regional domain | 4/26 | 15.4% | 4/4 |
| singapore-sglt2 | 2026-10-11 | ChatGPT | P45 | consumer / retail (incl. pharmacy) page, non-regional domain | 3/26 | 11.5% | 3/3 |
| uae-glp1 | 2026-09-29 | Perplexity | P50 | regulator / government health body page, UAE domain (public: doh.gov.ae) | 7/60 | 11.7% | 1/7 |
| uae-glp1 | 2026-09-29 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 12/60 | 20.0% | 10/12 |
| uae-glp1 | 2026-10-04 | Perplexity | P62 | consumer / retail (incl. pharmacy) page, non-regional domain | 15/60 | 25.0% | 8/15 |
| uae-glp1 | 2026-10-04 | Perplexity | P50 | regulator / government health body page, UAE domain (public: doh.gov.ae) | 9/60 | 15.0% | 2/9 |
| uae-glp1 | 2026-10-04 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 22/60 | 36.7% | 20/22 |
| uae-glp1 | 2026-10-06 | Perplexity | P62 | consumer / retail (incl. pharmacy) page, non-regional domain | 13/60 | 21.7% | 2/13 |
| uae-glp1 | 2026-10-06 | Perplexity | P50 | regulator / government health body page, UAE domain (public: doh.gov.ae) | 9/60 | 15.0% | 3/9 |
| uae-glp1 | 2026-10-06 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 23/60 | 38.3% | 23/23 |
| uae-glp1 | 2026-10-07 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 24/60 | 40.0% | 24/24 |
| uae-glp1 | 2026-10-08 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 21/60 | 35.0% | 21/21 |
| uae-glp1 | 2026-10-09 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 20/60 | 33.3% | 18/20 |
| uae-glp1 | 2026-10-10 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 23/60 | 38.3% | 20/23 |
| uae-glp1 | 2026-10-11 | ChatGPT | P85 | consumer / retail (incl. pharmacy) page, non-regional domain | 6/60 | 10.0% | 0/6 |
| uae-glp1 | 2026-10-11 | ChatGPT | P86 | consumer / retail (incl. pharmacy) page, non-regional domain | 6/60 | 10.0% | 0/6 |
| uae-glp1 | 2026-10-11 | Claude | P61 | a local commercial blog post (UAE), August 2026, before the congress | 19/60 | 31.7% | 17/19 |

**Gemini** (summary; every page-cycle is in the JSON). Gemini's pre-congress redirect links were never resolved, so a page Gemini cited before the congress also counts as new here (D16): Gemini's rows over-include and are not comparable with the other engines'.

| Series | Cycle date | Pages listed | Largest share | Types of the listed pages |
|---|---|---|---|---|
| saudi-t1d | 2026-10-04 | 15 | 50.0% | consumer / retail (incl. pharmacy) page, non-regional domain ×11; consumer / retail (incl. pharmacy) page, Saudi domain ×4 |
| saudi-t1d | 2026-10-05 | 10 | 44.6% | consumer / retail (incl. pharmacy) page, non-regional domain ×7; consumer / retail (incl. pharmacy) page, Saudi domain ×3 |
| saudi-t1d | 2026-10-06 | 12 | 46.4% | consumer / retail (incl. pharmacy) page, non-regional domain ×9; consumer / retail (incl. pharmacy) page, Saudi domain ×3 |
| saudi-t1d | 2026-10-07 | 10 | 48.2% | consumer / retail (incl. pharmacy) page, non-regional domain ×8; consumer / retail (incl. pharmacy) page, Saudi domain ×2 |
| saudi-t1d | 2026-10-08 | 12 | 50.0% | consumer / retail (incl. pharmacy) page, non-regional domain ×9; consumer / retail (incl. pharmacy) page, Saudi domain ×3 |
| saudi-t1d | 2026-10-09 | 11 | 48.2% | consumer / retail (incl. pharmacy) page, non-regional domain ×8; consumer / retail (incl. pharmacy) page, Saudi domain ×2; clinic / hospital / care provider page, non-regional domain ×1 |
| saudi-t1d | 2026-10-10 | 11 | 60.7% | consumer / retail (incl. pharmacy) page, non-regional domain ×8; consumer / retail (incl. pharmacy) page, Saudi domain ×3 |
| saudi-t1d | 2026-10-11 | 11 | 53.6% | consumer / retail (incl. pharmacy) page, non-regional domain ×8; consumer / retail (incl. pharmacy) page, Saudi domain ×3 |
| singapore-sglt2 | 2026-10-02 | 7 | 46.2% | consumer / retail (incl. pharmacy) page, non-regional domain ×4; clinic / hospital / care provider page, non-regional domain ×2; clinic / hospital / care provider page, Singapore domain ×1 |
| singapore-sglt2 | 2026-10-05 | 8 | 46.2% | clinic / hospital / care provider page, non-regional domain ×4; consumer / retail (incl. pharmacy) page, non-regional domain ×3; regulator / government health body page, Singapore domain ×1 |
| singapore-sglt2 | 2026-10-06 | 5 | 23.1% | clinic / hospital / care provider page, non-regional domain ×3; consumer / retail (incl. pharmacy) page, non-regional domain ×2 |
| singapore-sglt2 | 2026-10-07 | 8 | 50.0% | consumer / retail (incl. pharmacy) page, non-regional domain ×5; clinic / hospital / care provider page, non-regional domain ×3 |
| singapore-sglt2 | 2026-10-08 | 6 | 38.5% | clinic / hospital / care provider page, non-regional domain ×3; consumer / retail (incl. pharmacy) page, non-regional domain ×2; clinic / hospital / care provider page, Singapore domain ×1 |
| singapore-sglt2 | 2026-10-09 | 6 | 30.8% | clinic / hospital / care provider page, non-regional domain ×3; consumer / retail (incl. pharmacy) page, non-regional domain ×2; clinic / hospital / care provider page, Singapore domain ×1 |
| singapore-sglt2 | 2026-10-10 | 9 | 53.8% | clinic / hospital / care provider page, non-regional domain ×6; consumer / retail (incl. pharmacy) page, non-regional domain ×2; clinic / hospital / care provider page, Singapore domain ×1 |
| singapore-sglt2 | 2026-10-11 | 6 | 50.0% | clinic / hospital / care provider page, non-regional domain ×3; consumer / retail (incl. pharmacy) page, non-regional domain ×2; clinic / hospital / care provider page, Singapore domain ×1 |
| uae-glp1 | 2026-09-29 | 10 | 20.0% | clinic / hospital / care provider page, non-regional domain ×3; clinic / hospital / care provider page, UAE domain ×2; consumer / retail (incl. pharmacy) page, UAE domain ×2; consumer / retail (incl. pharmacy) page, non-regional domain ×2; unclassified page, UAE domain ×1 |
| uae-glp1 | 2026-10-04 | 13 | 21.7% | clinic / hospital / care provider page, UAE domain ×5; clinic / hospital / care provider page, non-regional domain ×4; consumer / retail (incl. pharmacy) page, non-regional domain ×2; consumer / retail (incl. pharmacy) page, UAE domain ×1; unclassified page, UAE domain ×1 |
| uae-glp1 | 2026-10-06 | 10 | 25.0% | consumer / retail (incl. pharmacy) page, non-regional domain ×4; clinic / hospital / care provider page, UAE domain ×2; consumer / retail (incl. pharmacy) page, UAE domain ×2; clinic / hospital / care provider page, non-regional domain ×1; insurer page, UAE domain ×1 |
| uae-glp1 | 2026-10-07 | 7 | 20.0% | clinic / hospital / care provider page, UAE domain ×3; consumer / retail (incl. pharmacy) page, UAE domain ×2; consumer / retail (incl. pharmacy) page, non-regional domain ×1; clinic / hospital / care provider page, non-regional domain ×1 |
| uae-glp1 | 2026-10-08 | 9 | 18.3% | clinic / hospital / care provider page, UAE domain ×4; clinic / hospital / care provider page, non-regional domain ×2; consumer / retail (incl. pharmacy) page, non-regional domain ×2; consumer / retail (incl. pharmacy) page, UAE domain ×1 |
| uae-glp1 | 2026-10-09 | 8 | 23.3% | consumer / retail (incl. pharmacy) page, non-regional domain ×3; clinic / hospital / care provider page, UAE domain ×3; clinic / hospital / care provider page, non-regional domain ×1; insurer page, UAE domain ×1 |
| uae-glp1 | 2026-10-10 | 10 | 13.3% | clinic / hospital / care provider page, non-regional domain ×5; consumer / retail (incl. pharmacy) page, non-regional domain ×2; clinic / hospital / care provider page, UAE domain ×2; consumer / retail (incl. pharmacy) page, UAE domain ×1 |
| uae-glp1 | 2026-10-11 | 7 | 18.3% | clinic / hospital / care provider page, non-regional domain ×2; clinic / hospital / care provider page, UAE domain ×2; consumer / retail (incl. pharmacy) page, non-regional domain ×1; unclassified page, UAE domain ×1; consumer / retail (incl. pharmacy) page, UAE domain ×1 |

**Claude, UAE: page P61.** A page new to the series (never cited in any before-window cycle, including the 27 Sep v5 cycle) first drew 12/60 of Claude's answers on 2026-09-29 and 22/60, 23/60, 24/60, 21/60, 20/60, 23/60, 19/60 in the after-window cycles; 153 of the 164 citing answers across these cycles name Wegovy. Its type, from the page's own content: a local commercial blog post (UAE), August 2026, before the congress. The type and date come from a read-only fetch on 11 Oct; item 12 tests the page's content at citation by echo. It is never named.

**What this shows.** Which newly cited local or regional pages carried a sizeable share of one engine's answers after the congress (UAE distinct pages listed: Perplexity 2, Claude 1, ChatGPT 2, Gemini 32 with over-inclusion). In the UAE, Claude's after-window citations concentrate on one new local page that entered on 29 Sep, the cycle in which its naming first rose. **What it does not show.** That any page moved naming (naming among citing answers is high partly because the page discusses the medicine); that a page said then what it says now; or anything about the publishers, who are never named.

### 1(c) Source-class mix per engine per window

Unit: distinct cited links per answer. Classifier: rule-based, by host, written in the script (named public entities; hand-assigned local businesses held as hashes; government and academic suffixes; host tokens). 4408 distinct hosts and 151614 links classified; **unclassified share 5.4%** of links. "Other" = reference publishers, professional or patient organisations, academic hosts, non-health government, market research, insurers, directories and social platforms (link counts by subclass in the JSON). Gemini is classified by the domain its API returns as each result's title.

**UAE, Wegovy (registered)** — share of links by class (links)

| Engine | Window | Answers | Regulator / government health body | Journal / literature / trial registry | Congress | Sponsor / manufacturer | News / medical media | Clinic / hospital / care provider | Consumer / retail (incl. pharmacy) | Other (classified) | Unclassified |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Perplexity | before-v4 | 765 | 8.9% (1303) | 5.7% (828) | 0.0% (0) | 0.0% (2) | 11.3% (1642) | 32.7% (4762) | 28.3% (4121) | 9.9% (1439) | 3.3% (477) |
| Perplexity | before-v5 | 58 | 8.4% (94) | 4.3% (48) | 0.0% (0) | 0.0% (0) | 14.9% (168) | 30.5% (343) | 29.3% (329) | 8.8% (99) | 3.8% (43) |
| Perplexity | during | 60 | 9.4% (112) | 4.6% (54) | 0.0% (0) | 0.0% (0) | 17.9% (212) | 26.8% (318) | 28.6% (339) | 9.7% (115) | 2.9% (35) |
| Perplexity | after | 420 | 6.7% (361) | 3.9% (211) | 0.0% (0) | 0.1% (3) | 12.7% (683) | 21.6% (1158) | 19.0% (1018) | 33.0% (1770) | 3.1% (165) |
| Gemini | before-v4 | 779 | 4.9% (1030) | 0.6% (125) | 0.0% (0) | 0.5% (105) | 10.2% (2143) | 40.7% (8510) | 25.2% (5268) | 9.6% (2017) | 8.3% (1726) |
| Gemini | before-v5 | 60 | 3.8% (63) | 0.6% (10) | 0.0% (0) | 0.5% (8) | 9.6% (159) | 41.7% (689) | 24.6% (406) | 10.6% (176) | 8.6% (143) |
| Gemini | during | 60 | 3.5% (54) | 0.4% (6) | 0.0% (0) | 0.5% (8) | 9.6% (149) | 43.9% (683) | 24.3% (378) | 9.4% (146) | 8.5% (132) |
| Gemini | after | 420 | 4.5% (507) | 0.5% (53) | 0.0% (0) | 0.7% (81) | 10.1% (1132) | 40.4% (4532) | 24.2% (2719) | 10.8% (1210) | 8.9% (996) |
| ChatGPT | before-v4 | 780 | 2.5% (96) | 0.2% (9) | 0.0% (0) | 0.2% (9) | 8.8% (338) | 24.3% (937) | 42.9% (1653) | 11.1% (427) | 9.9% (383) |
| ChatGPT | before-v5 | 60 | 0.3% (1) | 0.0% (0) | 0.0% (0) | 0.0% (0) | 8.1% (24) | 30.9% (91) | 44.4% (131) | 10.2% (30) | 6.1% (18) |
| ChatGPT | during | 60 | 2.0% (6) | 0.7% (2) | 0.0% (0) | 0.3% (1) | 7.4% (22) | 25.5% (76) | 44.3% (132) | 10.7% (32) | 9.1% (27) |
| ChatGPT | after | 420 | 1.8% (38) | 0.2% (5) | 0.0% (0) | 0.3% (7) | 9.6% (199) | 22.4% (465) | 45.8% (949) | 11.5% (239) | 8.3% (171) |
| Claude | before-v4 | 780 | 2.3% (140) | 5.9% (358) | 0.0% (0) | 0.1% (4) | 16.1% (978) | 31.5% (1916) | 22.2% (1350) | 15.2% (922) | 6.7% (406) |
| Claude | before-v5 | 60 | 2.1% (10) | 5.1% (24) | 0.0% (0) | 0.0% (0) | 16.6% (78) | 24.5% (115) | 26.7% (125) | 18.3% (86) | 6.6% (31) |
| Claude | during | 60 | 2.5% (12) | 5.9% (28) | 0.0% (0) | 0.0% (0) | 14.7% (70) | 26.0% (124) | 24.9% (119) | 18.4% (88) | 7.5% (36) |
| Claude | after | 420 | 2.2% (72) | 6.8% (220) | 0.0% (0) | 0.0% (0) | 14.5% (472) | 26.2% (851) | 29.4% (957) | 16.3% (529) | 4.7% (153) |

**Singapore, Jardiance (registered)** — share of links by class (links)

| Engine | Window | Answers | Regulator / government health body | Journal / literature / trial registry | Congress | Sponsor / manufacturer | News / medical media | Clinic / hospital / care provider | Consumer / retail (incl. pharmacy) | Other (classified) | Unclassified |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Perplexity | before-v4 | 385 | 25.3% (1771) | 10.3% (723) | 0.0% (0) | 0.8% (54) | 2.2% (153) | 34.2% (2391) | 10.8% (753) | 15.2% (1065) | 1.3% (89) |
| Perplexity | during | 26 | 19.8% (102) | 8.3% (43) | 0.0% (0) | 1.2% (6) | 16.1% (83) | 24.0% (124) | 8.7% (45) | 14.7% (76) | 7.2% (37) |
| Perplexity | after | 182 | 30.9% (689) | 6.6% (148) | 0.0% (0) | 0.8% (17) | 7.6% (170) | 29.0% (647) | 7.6% (170) | 13.0% (289) | 4.4% (98) |
| Gemini | before-v4 | 389 | 15.4% (956) | 3.4% (214) | 0.0% (0) | 2.3% (140) | 3.5% (216) | 37.6% (2344) | 8.6% (536) | 25.9% (1614) | 3.3% (205) |
| Gemini | during | 26 | 11.2% (54) | 1.0% (5) | 0.0% (0) | 1.7% (8) | 5.8% (28) | 32.0% (154) | 11.8% (57) | 34.1% (164) | 2.3% (11) |
| Gemini | after | 182 | 13.6% (434) | 2.4% (77) | 0.0% (0) | 1.0% (32) | 4.5% (143) | 36.5% (1164) | 9.4% (300) | 29.7% (946) | 2.8% (90) |
| ChatGPT | before-v4 | 390 | 62.8% (997) | 0.3% (4) | 0.0% (0) | 0.0% (0) | 0.7% (11) | 7.6% (120) | 22.2% (353) | 3.5% (55) | 3.0% (47) |
| ChatGPT | during | 26 | 61.5% (75) | 0.0% (0) | 0.0% (0) | 0.0% (0) | 3.3% (4) | 7.4% (9) | 23.8% (29) | 3.3% (4) | 0.8% (1) |
| ChatGPT | after | 182 | 62.9% (461) | 0.3% (2) | 0.0% (0) | 0.0% (0) | 0.3% (2) | 7.1% (52) | 21.0% (154) | 2.9% (21) | 5.6% (41) |
| Claude | before-v4 | 390 | 23.1% (465) | 8.8% (178) | 0.0% (0) | 1.1% (23) | 0.5% (11) | 13.1% (263) | 25.2% (508) | 24.7% (497) | 3.5% (70) |
| Claude | during | 26 | 17.0% (20) | 6.8% (8) | 0.0% (0) | 0.9% (1) | 0.0% (0) | 21.2% (25) | 28.0% (33) | 20.3% (24) | 5.9% (7) |
| Claude | after | 182 | 27.7% (222) | 7.9% (63) | 0.0% (0) | 0.9% (7) | 0.1% (1) | 20.1% (161) | 16.9% (135) | 21.5% (172) | 5.0% (40) |

**Saudi Arabia, Tzield (post-registration, D10)** — share of links by class (links)

| Engine | Window | Answers | Regulator / government health body | Journal / literature / trial registry | Congress | Sponsor / manufacturer | News / medical media | Clinic / hospital / care provider | Consumer / retail (incl. pharmacy) | Other (classified) | Unclassified |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Perplexity | before-v4 | 600 | 18.1% (2077) | 43.0% (4938) | 0.1% (9) | 7.3% (836) | 6.5% (740) | 5.9% (681) | 1.6% (184) | 14.1% (1624) | 3.4% (389) |
| Perplexity | after | 448 | 19.9% (1194) | 32.7% (1958) | 0.0% (0) | 5.9% (356) | 8.6% (515) | 2.6% (155) | 0.4% (27) | 26.9% (1611) | 2.9% (177) |
| Gemini | before-v4 | 614 | 13.2% (1844) | 6.5% (903) | 0.0% (6) | 8.6% (1209) | 5.0% (692) | 8.8% (1227) | 29.3% (4094) | 22.6% (3165) | 6.1% (847) |
| Gemini | after | 448 | 12.3% (1240) | 4.8% (486) | 0.0% (4) | 7.7% (779) | 4.8% (481) | 9.6% (971) | 34.5% (3492) | 20.3% (2053) | 6.0% (609) |
| ChatGPT | before-v4 | 616 | 10.2% (183) | 16.9% (303) | 0.0% (0) | 2.4% (42) | 4.5% (80) | 8.4% (151) | 37.3% (668) | 10.4% (187) | 9.8% (175) |
| ChatGPT | after | 448 | 13.2% (171) | 17.6% (228) | 0.0% (0) | 1.5% (19) | 2.7% (35) | 6.8% (88) | 40.1% (518) | 10.0% (129) | 8.1% (104) |
| Claude | before-v4 | 616 | 1.4% (61) | 29.8% (1321) | 0.0% (1) | 1.6% (71) | 9.6% (425) | 2.8% (123) | 3.1% (136) | 48.1% (2131) | 3.7% (164) |
| Claude | after | 445 | 1.6% (50) | 33.0% (1024) | 0.0% (0) | 2.2% (69) | 9.1% (284) | 3.7% (114) | 4.0% (125) | 43.5% (1353) | 2.9% (89) |

#### Claude, UAE: 27 Sep (v5, before) against 29 Sep (congress window)

Governed naming: 22/54 (40.7%) on 2026-09-27, 31/54 (57.4%) on 2026-09-29. Answers with at least one citation: 57/60 and 57/60. Distinct domains: 173 and 175 (64 new in the later cycle, 62 absent from it).

| Class | Earlier: links (share) | Later: links (share) | Change (pp) | Earlier: answers citing (naming) | Later: answers citing (naming) |
|---|---|---|---|---|---|
| Regulator / government health body | 10 (2.1%) | 12 (2.5%) | +0.4 | 9 (5) | 11 (8) |
| Journal / literature / trial registry | 24 (5.1%) | 28 (5.9%) | +0.8 | 12 (4) | 14 (6) |
| Congress | 0 (0.0%) | 0 (0.0%) | 0.0 | 0 (0) | 0 (0) |
| Sponsor / manufacturer | 0 (0.0%) | 0 (0.0%) | 0.0 | 0 (0) | 0 (0) |
| News / medical media | 78 (16.6%) | 70 (14.7%) | -1.9 | 35 (26) | 31 (24) |
| Clinic / hospital / care provider | 115 (24.5%) | 124 (26.0%) | +1.5 | 39 (25) | 40 (30) |
| Consumer / retail (incl. pharmacy) | 125 (26.7%) | 119 (24.9%) | -1.7 | 44 (18) | 42 (23) |
| Other (classified) | 86 (18.3%) | 88 (18.4%) | +0.1 | 43 (17) | 44 (27) |
| Unclassified | 31 (6.6%) | 36 (7.5%) | +0.9 | 21 (9) | 20 (9) |

Largest class change: News / medical media -1.9pp. Among the top new domains, answers citing by type: consumer / retail (incl. pharmacy) 17; news / medical media 4; unclassified 3; regulator / government health body 2; market research 2; clinic / hospital / care provider 1.

Top new domains in the later cycle (public entities named; others typed with an opaque id): [consumer / retail (incl. pharmacy), UAE domain, H001] — 12 answers (10 naming); [consumer / retail (incl. pharmacy), non-regional domain, H002] — 2 answers (1 naming); emedicine.medscape.com — 2 answers (1 naming); ods.od.nih.gov — 2 answers (1 naming); [consumer / retail (incl. pharmacy), UAE domain, H003] — 2 answers (2 naming); [market research, non-regional domain, H004] — 2 answers (1 naming); biospace.com — 2 answers (2 naming); [clinic / hospital / care provider, non-regional domain, H005] — 1 answers (1 naming); [unclassified, UAE domain, H006] — 1 answers (0 naming); [unclassified, UAE domain, H007] — 1 answers (0 naming); [consumer / retail (incl. pharmacy), non-regional domain, H008] — 1 answers (1 naming); [unclassified, non-regional domain, H009] — 1 answers (1 naming).

Most-cited domains absent from the later cycle: [market research, non-regional domain, H010] — 4 answers; [clinic / hospital / care provider, non-regional domain, H011] — 2 answers; [consumer / retail (incl. pharmacy), UAE domain, H012] — 2 answers; [consumer / retail (incl. pharmacy), non-regional domain, H013] — 1 answers; [unclassified, non-regional domain, H014] — 1 answers; [consumer / retail (incl. pharmacy), non-regional domain, H015] — 1 answers; ovid.com — 1 answers; [clinic / hospital / care provider, non-regional domain, H016] — 1 answers.

#### Perplexity, UAE: 6 Oct against 7 Oct (after window)

Governed naming: 17/54 (31.5%) on 2026-10-06, 40/54 (74.1%) on 2026-10-07. Answers with at least one citation: 60/60 and 60/60. Distinct domains: 199 and 115 (46 new in the later cycle, 130 absent from it).

| Class | Earlier: links (share) | Later: links (share) | Change (pp) | Earlier: answers citing (naming) | Later: answers citing (naming) |
|---|---|---|---|---|---|
| Regulator / government health body | 98 (8.3%) | 32 (5.3%) | -3.0 | 28 (11) | 17 (15) |
| Journal / literature / trial registry | 78 (6.6%) | 18 (3.0%) | -3.6 | 23 (8) | 12 (9) |
| Congress | 0 (0.0%) | 0 (0.0%) | 0.0 | 0 (0) | 0 (0) |
| Sponsor / manufacturer | 0 (0.0%) | 0 (0.0%) | 0.0 | 0 (0) | 0 (0) |
| News / medical media | 207 (17.5%) | 45 (7.5%) | -10.0 | 37 (20) | 25 (20) |
| Clinic / hospital / care provider | 301 (25.5%) | 116 (19.3%) | -6.2 | 46 (20) | 35 (33) |
| Consumer / retail (incl. pharmacy) | 339 (28.7%) | 68 (11.3%) | -17.4 | 38 (14) | 34 (24) |
| Other (classified) | 118 (10.0%) | 302 (50.3%) | +40.3 | 45 (17) | 56 (42) |
| Unclassified | 40 (3.4%) | 19 (3.2%) | -0.2 | 15 (8) | 14 (11) |

Largest class change: Other (classified) +40.3pp. Among the top new domains, answers citing by type: market research 63; consumer / retail (incl. pharmacy) 7; other (public) 6; journal / literature / trial registry 3; unclassified 2; clinic / hospital / care provider 2; directory or platform 2.

Top new domains in the later cycle (public entities named; others typed with an opaque id): [market research, non-regional domain, H017] — 25 answers (16 naming); [market research, non-regional domain, H018] — 14 answers (6 naming); [market research, non-regional domain, H019] — 11 answers (8 naming); [market research, non-regional domain, H020] — 7 answers (4 naming); [market research, non-regional domain, H021] — 6 answers (5 naming); drugs.com — 6 answers (6 naming); [consumer / retail (incl. pharmacy), UAE domain, H022] — 5 answers (3 naming); pubmed.ncbi.nlm.nih.gov — 3 answers (1 naming); [unclassified, non-regional domain, H023] — 2 answers (2 naming); [clinic / hospital / care provider, non-regional domain, H024] — 2 answers (2 naming); [directory or platform, non-regional domain, H025] — 2 answers (2 naming); [consumer / retail (incl. pharmacy), non-regional domain, H026] — 2 answers (2 naming).

Most-cited domains absent from the later cycle: [consumer / retail (incl. pharmacy), UAE domain, H027] — 21 answers; [consumer / retail (incl. pharmacy), non-regional domain, H028] — 19 answers; [consumer / retail (incl. pharmacy), non-regional domain, H029] — 13 answers; [clinic / hospital / care provider, UAE domain, H030] — 13 answers; zawya.com — 9 answers; [clinic / hospital / care provider, UAE domain, H031] — 9 answers; globenewswire.com — 8 answers; [clinic / hospital / care provider, UAE domain, H032] — 8 answers.

**What this shows.** The kinds of source each engine cited in each window, and what changed in the two cycle pairs where naming stepped. **What it does not show.** That a class change caused a naming change (both windows share questions, and the engine's own retrieval and model can change on the same day); that a cited page said what the class implies; or anything about engines' consumer products. Host-level classes are approximate: a host's class is its publisher's type, not the page's content.

## 2. E2 band-rule readings

Protocol §4: the band is the v4 single-cycle range "shifted by the §6 v4→v5 offset for that engine, with each end widened by the offset's 95% interval". The registered decision uses reading (i), I1 (D16). (ii) and (iii) are the other readings the methodologist raised (M1); a shift-only band, which omits the registered widening, is shown for reference. The after-window rate is the registered pooled rate; bands use the registered limits and stored offset bounds, as the decision did. Signed distance: positive = beyond the nearer edge, negative = inside by that much.

| Series | Engine | After K/N (rate) | Registered verdict | (i) I1 band → verdict (distance pp) | (ii) half-width → verdict (distance) | (iii) full width → verdict (distance) | Shift only (reference) | Flips under (ii)/(iii) |
|---|---|---|---|---|---|---|---|---|
| uae-glp1 | Perplexity | 243/378 (64.286%) | no movement | 41.98%–64.36% → no movement (-0.074) | 41.36%–63.74% → **movement** (+0.546) | 36.42%–68.68% → no movement (-4.394) | 46.30%–58.80% → **movement** (+5.486) | ii-shift-widen-half-width |
| uae-glp1 | Gemini | 206/378 (54.497%) | no movement | 35.14%–78.42% → no movement (-19.357) | 35.14%–78.42% → no movement (-19.357) | 26.50%–87.06% → no movement (-27.997) | 43.78%–69.78% → no movement (-10.717) | none |
| uae-glp1 | ChatGPT | 110/378 (29.101%) | no movement | 14.23%–46.87% → no movement (-14.870) | 14.23%–46.87% → no movement (-14.870) | 9.91%–51.19% → no movement (-19.191) | 18.55%–42.55% → no movement (-10.550) | none |
| uae-glp1 | Claude | 256/378 (67.725%) | movement | 34.53%–67.33% → **movement** (+0.395) | 34.84%–67.64% → **movement** (+0.085) | 25.89%–76.59% → no movement (-8.865) | 43.79%–58.69% → **movement** (+9.035) | iii-shift-widen-full-width |
| uae-glp1 | Four engines | 815/1512 (53.902%) | no movement | 36.88%–53.98% → no movement (-0.078) | 36.88%–53.98% → no movement (-0.078) | 33.33%–57.53% → no movement (-3.628) | 40.43%–50.43% → **movement** (+3.472) | none |
| saudi-t1d (post-reg.) | Perplexity | 26/440 (5.909%) | no movement | 3.77%–23.64% → no movement (-2.135) | 3.47%–23.34% → no movement (-2.436) | 2.56%–24.25% → no movement (-3.345) | 4.38%–22.43% → no movement (-1.526) | none |
| saudi-t1d (post-reg.) | Gemini | 40/440 (9.091%) | no movement | 3.03%–30.30% → no movement (-6.064) | 3.03%–30.30% → no movement (-6.059) | -2.42%–35.76% → no movement (-11.514) | 8.49%–24.85% → no movement (-0.605) | none |
| saudi-t1d (post-reg.) | ChatGPT | 81/440 (18.409%) | no movement | 6.67%–35.15% → no movement (-11.740) | 6.67%–35.15% → no movement (-11.740) | 2.43%–39.39% → no movement (-15.980) | 10.91%–30.91% → no movement (-7.500) | none |
| saudi-t1d (post-reg.) | Claude | 28/437 (6.407%) | no movement | -1.21%–13.33% → no movement (-6.922) | -1.21%–13.33% → no movement (-6.927) | -3.02%–15.15% → no movement (-8.742) | 0.61%–11.52% → no movement (-5.112) | none |
| saudi-t1d (post-reg.) | Four engines | 175/1757 (9.960%) | no movement | 6.49%–23.33% → no movement (-3.466) | 6.57%–23.41% → no movement (-3.386) | 4.68%–25.30% → no movement (-5.276) | 8.46%–21.52% → no movement (-1.496) | none |

**What this shows.** Which verdicts depend on the reading. uae-glp1 Perplexity (no movement under I1) flips under ii-shift-widen-half-width; uae-glp1 Claude (movement under I1) flips under iii-shift-widen-full-width. **What it does not show.** Which reading the author intended: I1 was fixed in D16 after per-cycle counts had been displayed. The registered verdicts are those of I1 and stand; this table never replaces them.

## 3. Cycle-level change estimate

mean after-window cycle rate minus mean v4 reference cycle rate minus the bridge point offset; cycles are the unit. A different estimand from the band rule. SE = sqrt(sd_after²/n_after + sd_before²/n_before + se_offset²); se_offset = (upper − lower)/3.92 from the bootstrap 95% interval (normal approximation); conservative variant uses max(offset − lower, upper − offset)/1.96. z and 95% CI by normal approximation; cycles treated as independent.

| Series | Engine | Before: cycles, mean (SD) | After: cycles, mean (SD) | Offset (95% CI) | Change net of offset | SE (cycles / offset / total) | z | 95% CI | z with conservative offset SE |
|---|---|---|---|---|---|---|---|---|---|
| uae-glp1 | Perplexity | 7, 52.1% (3.93pp) | 7, 64.3% (19.03pp) | 0.00 (-4.32 to +5.56) | **+12.19pp** | 7.34 / 2.52 / 7.76 | 1.57 | -3.0 to +27.4 | 1.55 |
| uae-glp1 | Gemini | 13, 59.3% (7.49pp) | 7, 54.5% (5.84pp) | -4.32 (-12.96 to +4.32) | **-0.51pp** | 3.03 / 4.41 / 5.35 | -0.10 | -11.0 to +10.0 | -0.10 |
| uae-glp1 | ChatGPT | 13, 30.2% (5.97pp) | 7, 29.1% (4.87pp) | -1.85 (-6.17 to +2.47) | **+0.75pp** | 2.48 / 2.20 / 3.31 | 0.23 | -5.7 to +7.2 | 0.23 |
| uae-glp1 | Claude | 13, 47.9% (4.40pp) | 7, 67.7% (3.36pp) | +3.09 (-6.17 to +11.73) | **+16.77pp** | 1.76 / 4.57 / 4.89 | 3.43 | +7.2 to +26.4 | 3.33 |
| uae-glp1 | Four engines | 13, 45.6% (2.87pp) | 7, 53.9% (4.72pp) | -0.77 (-4.32 to +2.78) | **+9.10pp** | 1.95 / 1.81 / 2.66 | 3.42 | +3.9 to +14.3 | 3.42 |
| saudi-t1d (post-reg.) | Perplexity | 11, 9.8% (5.59pp) | 8, 5.9% (7.70pp) | +0.61 (0.00 to +1.82) | **-4.49pp** | 3.20 / 0.46 / 3.24 | -1.39 | -10.8 to +1.9 | -1.38 |
| saudi-t1d (post-reg.) | Gemini | 11, 9.0% (5.17pp) | 8, 9.1% (2.92pp) | +4.85 (-0.61 to +10.30) | **-4.71pp** | 1.87 / 2.78 / 3.35 | -1.41 | -11.3 to +1.9 | -1.40 |
| saudi-t1d (post-reg.) | ChatGPT | 11, 17.4% (6.52pp) | 8, 18.4% (3.82pp) | 0.00 (-4.24 to +4.24) | **+1.05pp** | 2.39 / 2.16 / 3.22 | 0.33 | -5.3 to +7.4 | 0.33 |
| saudi-t1d (post-reg.) | Claude | 11, 5.0% (4.23pp) | 8, 6.4% (4.50pp) | +0.61 (-1.21 to +2.42) | **+0.86pp** | 2.04 / 0.93 / 2.24 | 0.38 | -3.5 to +5.3 | 0.38 |
| saudi-t1d (post-reg.) | Four engines | 11, 10.3% (4.32pp) | 8, 10.0% (2.73pp) | +1.52 (-0.45 to +3.33) | **-1.83pp** | 1.62 / 0.96 / 1.89 | -0.97 | -5.5 to +1.9 | -0.96 |

**What this shows.** Treating cycles as the unit, Claude's UAE change net of the offset is +16.8pp (95% CI +7.2 to +26.4, z = 3.43); Perplexity's is +12.2pp (-3.0 to +27.4, z = 1.57), its SE inflated by the after-window regime switch (SD 19.0pp). This is **a different estimand from the band rule**: it asks whether the mean level changed, not whether the pooled rate left a range widened by interval bounds. **What it does not show.** A registered verdict; attribution of any change to the congress; and it assumes independent cycles, a normal approximation, an additive offset on the percentage scale measured on one day, and a symmetric offset interval (the conservative column relaxes the last).

## 4. Per-cycle view against the bridged band

Each before-v5, congress-window and after-window cycle against the registered bridged band (I1). Wilson 95% intervals assume independent answers and are therefore too narrow (answers to one question are correlated). "Clear" = the whole Wilson interval lies outside the band. Descriptive; the registered rule pools the after window.

**UAE, Wegovy (registered)**

| Engine | Bridged band | 09-27 (before-v5) | 09-29 (during) | 10-04 (after) | 10-06 (after) | 10-07 (after) | 10-08 (after) | 10-09 (after) | 10-10 (after) | 10-11 (after) | After cycles outside (above / below) |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Perplexity | 41.98%–64.36% | 25/52 48.1% [35.1%–61.3%] **inside** | 22/54 40.7% [28.7%–54.0%] **below** +1.24 | 23/54 42.6% [30.3%–55.8%] **inside** | 17/54 31.5% [20.7%–44.7%] **below** +10.50 | 40/54 74.1% [61.1%–83.9%] **above** +9.71 | 41/54 75.9% [63.0%–85.4%] **above** +11.57 | 40/54 74.1% [61.1%–83.9%] **above** +9.71 | 39/54 72.2% [59.1%–82.4%] **above** +7.86 | 43/54 79.6% [67.1%–88.2%] **above** +15.27 (clear) | 6/7 (5 / 1) |
| Gemini | 35.14%–78.42% | 34/54 63.0% [49.6%–74.6%] **inside** | 35/54 64.8% [51.5%–76.2%] **inside** | 31/54 57.4% [44.2%–69.7%] **inside** | 32/54 59.3% [46.0%–71.3%] **inside** | 32/54 59.3% [46.0%–71.3%] **inside** | 30/54 55.6% [42.4%–68.0%] **inside** | 28/54 51.9% [38.9%–64.6%] **inside** | 30/54 55.6% [42.4%–68.0%] **inside** | 23/54 42.6% [30.3%–55.8%] **inside** | 0/7 (0 / 0) |
| ChatGPT | 14.23%–46.87% | 15/54 27.8% [17.6%–40.9%] **inside** | 15/54 27.8% [17.6%–40.9%] **inside** | 13/54 24.1% [14.6%–37.0%] **inside** | 16/54 29.6% [19.1%–42.8%] **inside** | 16/54 29.6% [19.1%–42.8%] **inside** | 14/54 25.9% [16.1%–38.9%] **inside** | 16/54 29.6% [19.1%–42.8%] **inside** | 21/54 38.9% [27.0%–52.2%] **inside** | 14/54 25.9% [16.1%–38.9%] **inside** | 0/7 (0 / 0) |
| Claude | 34.53%–67.33% | 22/54 40.7% [28.7%–54.0%] **inside** | 31/54 57.4% [44.2%–69.7%] **inside** | 36/54 66.7% [53.4%–77.8%] **inside** | 37/54 68.5% [55.3%–79.3%] **above** +1.19 | 40/54 74.1% [61.1%–83.9%] **above** +6.74 | 37/54 68.5% [55.3%–79.3%] **above** +1.19 | 36/54 66.7% [53.4%–77.8%] **inside** | 34/54 63.0% [49.6%–74.6%] **inside** | 36/54 66.7% [53.4%–77.8%] **inside** | 3/7 (3 / 0) |
| Four engines | 36.88%–53.98% | 96/214 44.9% [38.4%–51.6%] **inside** | 103/216 47.7% [41.1%–54.3%] **inside** | 103/216 47.7% [41.1%–54.3%] **inside** | 102/216 47.2% [40.7%–53.9%] **inside** | 128/216 59.3% [52.6%–65.6%] **above** +5.28 | 122/216 56.5% [49.8%–62.9%] **above** +2.50 | 120/216 55.6% [48.9%–62.0%] **above** +1.58 | 124/216 57.4% [50.7%–63.8%] **above** +3.43 | 116/216 53.7% [47.0%–60.2%] **inside** | 4/7 (4 / 0) |

**Saudi Arabia, Tzield (post-registration, D10)**

| Engine | Bridged band | 10-04 (after) | 10-05 (after) | 10-06 (after) | 10-07 (after) | 10-08 (after) | 10-09 (after) | 10-10 (after) | 10-11 (after) | After cycles outside (above / below) |
|---|---|---|---|---|---|---|---|---|---|---|
| Perplexity | 3.77%–23.64% | 5/55 9.1% [4.0%–19.6%] **inside** | 6/55 10.9% [5.1%–21.8%] **inside** | 12/55 21.8% [13.0%–34.4%] **inside** | 1/55 1.8% [0.3%–9.6%] **below** +1.95 | 0/55 0.0% [0.0%–6.5%] **below** +3.77 | 0/55 0.0% [0.0%–6.5%] **below** +3.77 | 2/55 3.6% [1.0%–12.3%] **below** +0.13 | 0/55 0.0% [0.0%–6.5%] **below** +3.77 | 5/8 (0 / 5) |
| Gemini | 3.03%–30.30% | 4/55 7.3% [2.9%–17.3%] **inside** | 8/55 14.5% [7.6%–26.2%] **inside** | 5/55 9.1% [4.0%–19.6%] **inside** | 6/55 10.9% [5.1%–21.8%] **inside** | 4/55 7.3% [2.9%–17.3%] **inside** | 4/55 7.3% [2.9%–17.3%] **inside** | 6/55 10.9% [5.1%–21.8%] **inside** | 3/55 5.5% [1.9%–14.8%] **inside** | 0/8 (0 / 0) |
| ChatGPT | 6.67%–35.15% | 9/55 16.4% [8.9%–28.3%] **inside** | 6/55 10.9% [5.1%–21.8%] **inside** | 12/55 21.8% [13.0%–34.4%] **inside** | 9/55 16.4% [8.9%–28.3%] **inside** | 10/55 18.2% [10.2%–30.3%] **inside** | 11/55 20.0% [11.6%–32.4%] **inside** | 12/55 21.8% [13.0%–34.4%] **inside** | 12/55 21.8% [13.0%–34.4%] **inside** | 0/8 (0 / 0) |
| Claude | -1.21%–13.33% | 0/55 0.0% [0.0%–6.5%] **inside** | 5/52 9.6% [4.2%–20.6%] **inside** | 5/55 9.1% [4.0%–19.6%] **inside** | 6/55 10.9% [5.1%–21.8%] **inside** | 0/55 0.0% [0.0%–6.5%] **inside** | 3/55 5.5% [1.9%–14.8%] **inside** | 3/55 5.5% [1.9%–14.8%] **inside** | 6/55 10.9% [5.1%–21.8%] **inside** | 0/8 (0 / 0) |
| Four engines | 6.49%–23.33% | 18/220 8.2% [5.2%–12.6%] **inside** | 25/217 11.5% [7.9%–16.4%] **inside** | 34/220 15.5% [11.3%–20.8%] **inside** | 22/220 10.0% [6.7%–14.7%] **inside** | 14/220 6.4% [3.8%–10.4%] **below** +0.13 | 18/220 8.2% [5.2%–12.6%] **inside** | 23/220 10.5% [7.1%–15.2%] **inside** | 21/220 9.5% [6.3%–14.1%] **inside** | 1/8 (0 / 1) |

**The 27 Sep UAE cycle (before-v5, admitted).** Claude named Wegovy in 22/54 answers (40.7%), inside its bridged band, and in 31/54 (57.4%) on 29 Sep (congress window, inside). Claude's rise therefore began between the 27 Sep and 29 Sep cycles, both on the v5 instrument.

**What this shows.** Per cycle, Claude's after-window naming lies outside its bridged band in 3 of 7 cycles (3 above); Perplexity's in 6 of 7 (5 above, 1 below): its pooled "inside" verdict averages a low segment and a high one. **What it does not show.** A registered verdict (the rule is pooled), or why either engine moved.

## 5. Corrected bridge self-check

Expected range of n independent draws ∝ d2(n). With no between-day component, E[R3]/E[Rk] = d2(3)/d2(k). d2-adjusted ratio = (R3/d2(3)) / (Rk/d2(k)): ≈ 1 when the same-day spread accounts for the band, < 1 when between-day variation exceeds it. SDs: between-cycle (k − 1 df) and within-day from the three same-day runs (2 df per arm; pooled v4+v5, 4 df). One-sided F test of between > within. Within-day SDs rest on 2 (or 4 pooled) degrees of freedom and are very imprecise. Singapore is at ceiling and is omitted. Band widths use the exact single-cycle range of the cycles that define each band (UAE Perplexity: the 7 cycles from 9 Sep).

| Series | Engine | k | Band width (pp) | v4 runs | v5 runs | Same-day spread v4 / v5 (pp) | Registered ratio (v4 spread / width) | Expected with no day effect, d2(3)/d2(k) | d2-adjusted ratio: v4 / v5 / mean | SD between cycles (pp) | SD within day v4 / v5 / pooled (pp) | F (pooled), p one-sided | Material excess (ratio < 0.5): v4 runs / v4+v5 runs | F p < 0.05 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| uae-glp1 | Perplexity | 7 | 12.53 | 19/54, 16/54, 16/54 | 18/54, 15/54, 18/54 | 5.56 / 5.56 | 0.444 | 0.626 | 0.708 / 0.708 / 0.708 | 3.93 | 3.21 / 3.21 / 3.21 | 1.50, 0.362 | no / no | no |
| uae-glp1 | Gemini | 13 | 25.93 | 31/54, 32/54, 30/54 | 31/54, 26/54, 29/54 | 3.70 / 9.26 | 0.143 | 0.507 | 0.281 / 0.704 / 0.493 | 7.49 | 1.85 / 4.66 / 3.55 | 4.46, 0.080 | yes / yes | no |
| uae-glp1 | ChatGPT | 13 | 24.07 | 17/54, 16/54, 15/54 | 16/54, 15/54, 14/54 | 3.70 / 3.70 | 0.154 | 0.507 | 0.303 / 0.303 / 0.303 | 5.97 | 1.85 / 1.85 / 1.85 | 10.40, 0.018 | yes / yes | yes |
| uae-glp1 | Claude | 13 | 14.81 | 37/54, 36/54, 33/54 | 36/54, 38/54, 37/54 | 7.41 / 3.70 | 0.500 | 0.507 | 0.985 / 0.493 / 0.739 | 4.40 | 3.85 / 1.85 / 3.02 | 2.12, 0.245 | no / no | no |
| uae-glp1 | Four engines | 13 | 9.97 | 104/216, 100/216, 94/216 | 101/216, 94/216, 98/216 | 4.63 / 3.24 | 0.464 | 0.507 | 0.915 / 0.640 / 0.778 | 2.87 | 2.33 / 1.63 / 2.01 | 2.04, 0.257 | no / no | no |
| saudi-t1d (post-reg.) | Perplexity | 11 | 18.04 | 0/55, 1/55, 0/55 | 0/55, 1/55, 1/55 | 1.82 / 1.82 | 0.101 | 0.534 | 0.189 / 0.189 / 0.189 | 5.59 | 1.05 / 1.05 / 1.05 | 28.38, 0.003 | yes / yes | yes |
| saudi-t1d (post-reg.) | Gemini | 11 | 16.36 | 5/55, 3/55, 3/55 | 9/55, 2/54, 8/55 | 3.64 / 12.66 | 0.222 | 0.534 | 0.416 / 1.450 / 0.933 | 5.17 | 2.10 / 6.84 / 5.06 | 1.04, 0.530 | yes / no | no |
| saudi-t1d (post-reg.) | ChatGPT | 11 | 20.00 | 11/55, 9/55, 8/55 | 10/55, 7/55, 11/55 | 5.45 / 7.27 | 0.273 | 0.534 | 0.511 / 0.681 / 0.596 | 6.52 | 2.78 / 3.78 / 3.32 | 3.86, 0.102 | no / no | no |
| saudi-t1d (post-reg.) | Claude | 11 | 10.91 | 0/55, 1/55, 2/55 | 1/55, 2/55, 1/55 | 3.64 / 1.82 | 0.333 | 0.534 | 0.625 / 0.312 / 0.469 | 4.23 | 1.82 / 1.05 / 1.48 | 8.13, 0.029 | no / yes | yes |
| saudi-t1d (post-reg.) | Four engines | 11 | 13.06 | 16/220, 14/220, 13/220 | 20/220, 12/219, 21/220 | 1.36 / 4.07 | 0.104 | 0.534 | 0.196 / 0.584 / 0.390 | 4.32 | 0.69 / 2.23 / 1.65 | 6.86, 0.039 | yes / yes | yes |

Reviewers' expectation (Gemini, ChatGPT and Saudi Perplexity show excess between-day variation; UAE Claude and Perplexity do not), computed rather than assumed:

| Series | Engine | Expected | Material on v4 runs (matches) | Material on v4+v5 runs (matches) | F test p < 0.05 |
|---|---|---|---|---|---|
| uae-glp1 | Gemini | excess | excess (yes) | excess (yes) | no |
| uae-glp1 | ChatGPT | excess | excess (yes) | excess (yes) | yes |
| saudi-t1d | Perplexity | excess | excess (yes) | excess (yes) | yes |
| uae-glp1 | Claude | no excess | no excess (yes) | no excess (yes) | no |
| uae-glp1 | Perplexity | no excess | no excess (yes) | no excess (yes) | no |

**What this shows.** "Material" excess is defined here as a between-day sigma at least twice the within-day sigma (d2-adjusted ratio < 0.5); the threshold was chosen after the ratios were seen, so the classification is descriptive, and the F test beside it needs no threshold. On the v4 runs, the registered comparison, it holds for UAE Gemini, UAE ChatGPT, Saudi Perplexity, Saudi Gemini; with the v5 runs included it holds for UAE Gemini, UAE ChatGPT, Saudi Perplexity, Saudi Claude; the F test (4 df within) reaches p < 0.05 for UAE ChatGPT, Saudi Perplexity, Saudi Claude. UAE Claude's v4 same-day spread is what the n-difference alone predicts (ratio 0.99), so for the engine carrying the declared movement the data show no excess between-day variation. The classification of the smaller engine-series moves with the run set, because three runs give a very imprecise within-day estimate. The registered check compared a 3-run range with a k-cycle range, which is not like with like; as registered it was ill-posed as well as failed. **What it does not show.** The day × setting interaction (one bridge day cannot identify it), or that the offset measured at the 5 Oct operating point transfers to the band period.

## 6. Brand-only naming beside the registered measure

**The registered definition, exactly.** E2's naming outcome is the production scorer's `brandMentioned` for the answer, with the scan's brand names: the brand, every alias configured on the scan input, and, where the scan's domain grants molecule credit, the molecule the domain resolves for the brand. The names used in every admitted cycle were:

| Series | Brand | Registered names (brand or molecule) | Molecule credit | Brand-only names (this item) |
|---|---|---|---|---|
| UAE, Wegovy (registered) | Wegovy | Wegovy, semaglutide | molecule credit on (semaglutide) | Wegovy |
| Singapore, Jardiance (registered) | Jardiance | Jardiance, empagliflozin | molecule credit off (alias list only) | Jardiance |
| Saudi Arabia, Tzield (post-registration, D10) | Tzield | Tzield, teplizumab | molecule credit off (alias list only) | Tzield |

So the registered measure counts an answer that names only the molecule. Brand-only naming reruns the same scorer with the molecule removed. No bridge offset exists for brand-only naming (the bridge measured the registered measure), so v5 windows against v4 windows are descriptive and unbridged.

**UAE, Wegovy (registered)** — governed answers; registered (brand or molecule) and brand-only, K/N (rate)

| Window | Perplexity | Gemini | ChatGPT | Claude | Four engines |
|---|---|---|---|---|---|
| before-v4 | 309/689 (44.8%) · brand-only 279 (40.5%) | 416/701 (59.3%) · brand-only 392 (55.9%) | 212/702 (30.2%) · brand-only 197 (28.1%) | 336/702 (47.9%) · brand-only 319 (45.4%) | 1273/2794 (45.6%) · brand-only 1187 (42.5%) |
| before-v5 | 25/52 (48.1%) · brand-only 21 (40.4%) | 34/54 (63.0%) · brand-only 34 (63.0%) | 15/54 (27.8%) · brand-only 12 (22.2%) | 22/54 (40.7%) · brand-only 21 (38.9%) | 96/214 (44.9%) · brand-only 88 (41.1%) |
| during | 22/54 (40.7%) · brand-only 18 (33.3%) | 35/54 (64.8%) · brand-only 31 (57.4%) | 15/54 (27.8%) · brand-only 14 (25.9%) | 31/54 (57.4%) · brand-only 29 (53.7%) | 103/216 (47.7%) · brand-only 92 (42.6%) |
| after | 243/378 (64.3%) · brand-only 196 (51.9%) | 206/378 (54.5%) · brand-only 199 (52.6%) | 110/378 (29.1%) · brand-only 103 (27.2%) | 256/378 (67.7%) · brand-only 254 (67.2%) | 815/1512 (53.9%) · brand-only 752 (49.7%) |

**Singapore, Jardiance (registered)** — governed answers; registered (brand or molecule) and brand-only, K/N (rate)

| Window | Perplexity | Gemini | ChatGPT | Claude | Four engines |
|---|---|---|---|---|---|
| before-v4 | 283/283 (100.0%) · brand-only 282 (99.6%) | 281/284 (98.9%) · brand-only 281 (98.9%) | 285/285 (100.0%) · brand-only 285 (100.0%) | 284/285 (99.6%) · brand-only 281 (98.6%) | 1133/1137 (99.6%) · brand-only 1129 (99.3%) |
| during | 19/19 (100.0%) · brand-only 19 (100.0%) | 19/19 (100.0%) · brand-only 19 (100.0%) | 19/19 (100.0%) · brand-only 19 (100.0%) | 19/19 (100.0%) · brand-only 19 (100.0%) | 76/76 (100.0%) · brand-only 76 (100.0%) |
| after | 133/133 (100.0%) · brand-only 133 (100.0%) | 133/133 (100.0%) · brand-only 133 (100.0%) | 133/133 (100.0%) · brand-only 133 (100.0%) | 133/133 (100.0%) · brand-only 133 (100.0%) | 532/532 (100.0%) · brand-only 532 (100.0%) |

**Saudi Arabia, Tzield (post-registration, D10)** — governed answers; registered (brand or molecule) and brand-only, K/N (rate)

| Window | Perplexity | Gemini | ChatGPT | Claude | Four engines |
|---|---|---|---|---|---|
| before-v4 | 58/589 (9.8%) · brand-only 27 (4.6%) | 54/603 (9.0%) · brand-only 36 (6.0%) | 105/605 (17.4%) · brand-only 105 (17.4%) | 30/605 (5.0%) · brand-only 28 (4.6%) | 247/2402 (10.3%) · brand-only 196 (8.2%) |
| after | 26/440 (5.9%) · brand-only 19 (4.3%) | 40/440 (9.1%) · brand-only 20 (4.5%) | 81/440 (18.4%) · brand-only 81 (18.4%) | 28/437 (6.4%) · brand-only 27 (6.2%) | 175/1757 (10.0%) · brand-only 147 (8.4%) |

**What this shows.** In the UAE, Claude named the brand itself in 45.4% of v4 answers and 67.2% of after-window answers (registered: 47.9% → 67.7%); Perplexity 40.5% → 51.9% (registered 44.8% → 64.3%; v4 here is all v4 cycles, not the post-step reference). Answers credited by the molecule alone are the gap between the two rates. Cross-checks: the scorer's own canonical-name flag disagrees with the brand-only rescoring in 0 governed answers, and 0 answers are brand-only-named but not registered-named. **What it does not show.** A bridged brand-only change, a brand-only band, or a verdict.

## 7. Search rate, and attribution among answers that searched

**What "searched" can mean, per API, in this instrument.** The pull stores, per answer, the cited links and the search results the request path kept. In every answer the two are empty together or non-empty together, so "at least one citation" and "any searchResults" coincide here. Neither records a search the model ran without citing anything, so both are lower bounds on search invocation.

| Engine | Endpoint as configured | What counts as a citation | What searchResults holds |
|---|---|---|---|
| ChatGPT | gpt-4o-mini, OpenAI Responses API with the web_search tool; the model decides per request whether to search | url_citation annotations on the final message, deduplicated by URL | the same cited URLs and titles (no snippets); uncited results are not returned |
| Claude | claude-haiku-4-5, Messages API with the web_search_20250305 server tool, max_uses 5; the model decides whether to search | web_search_result_location citations in text blocks, deduplicated by URL | the cited results only; retrieved-but-uncited results are read for dates and then discarded |
| Gemini | gemini-2.5-flash with Google Search grounding; grounding is model-decided | groundingChunks web URIs (Google-hosted redirect links, one per chunk) | the same chunks, with the source domain as the title |
| Perplexity | sonar, chat completions; retrieval on every request | the response's citations list (or search_results URLs) | the API's search_results |

**UAE, Wegovy (registered)** — answers with at least one citation / captured answers (rate); mean citations and distinct domains per citing answer

| Window | Perplexity | Gemini | ChatGPT | Claude |
|---|---|---|---|---|
| before-v4 | 765/765 (100.0%); 19.1 cit., 13.6 dom. | 770/779 (98.8%); 27.2 cit., 23.6 dom. | 521/780 (66.8%); 7.4 cit., 6.5 dom. | 754/780 (96.7%); 8.1 cit., 7.4 dom. |
| before-v5 | 58/58 (100.0%); 19.4 cit., 12.3 dom. | 60/60 (100.0%); 27.6 cit., 23.5 dom. | 39/60 (65.0%); 7.6 cit., 6.6 dom. | 57/60 (95.0%); 8.2 cit., 7.5 dom. |
| during | 60/60 (100.0%); 19.8 cit., 12.2 dom. | 60/60 (100.0%); 25.9 cit., 22.3 dom. | 40/60 (66.7%); 7.5 cit., 6.9 dom. | 57/60 (95.0%); 8.4 cit., 7.8 dom. |
| after | 420/420 (100.0%); 12.8 cit., 9.9 dom. | 418/420 (99.5%); 26.9 cit., 22.9 dom. | 278/420 (66.2%); 7.5 cit., 6.5 dom. | 405/420 (96.4%); 8.0 cit., 7.3 dom. |

**Singapore, Jardiance (registered)** — answers with at least one citation / captured answers (rate); mean citations and distinct domains per citing answer

| Window | Perplexity | Gemini | ChatGPT | Claude |
|---|---|---|---|---|
| before-v4 | 385/385 (100.0%); 18.2 cit., 13.2 dom. | 380/389 (97.7%); 16.4 cit., 12.0 dom. | 390/390 (100.0%); 4.1 cit., 3.2 dom. | 390/390 (100.0%); 5.2 cit., 4.6 dom. |
| during | 26/26 (100.0%); 19.8 cit., 13.2 dom. | 26/26 (100.0%); 18.5 cit., 13.3 dom. | 26/26 (100.0%); 4.7 cit., 3.6 dom. | 26/26 (100.0%); 4.5 cit., 4.1 dom. |
| after | 182/182 (100.0%); 12.2 cit., 8.4 dom. | 181/182 (99.5%); 17.6 cit., 12.4 dom. | 182/182 (100.0%); 4.0 cit., 3.2 dom. | 182/182 (100.0%); 4.4 cit., 4.1 dom. |

**Saudi Arabia, Tzield (post-registration, D10)** — answers with at least one citation / captured answers (rate); mean citations and distinct domains per citing answer

| Window | Perplexity | Gemini | ChatGPT | Claude |
|---|---|---|---|---|
| before-v4 | 600/600 (100.0%); 19.1 cit., 14.2 dom. | 589/614 (95.9%); 23.7 cit., 19.0 dom. | 372/616 (60.4%); 4.8 cit., 4.2 dom. | 598/616 (97.1%); 7.4 cit., 6.6 dom. |
| after | 448/448 (100.0%); 13.4 cit., 10.8 dom. | 436/448 (97.3%); 23.2 cit., 18.6 dom. | 261/448 (58.3%); 5.0 cit., 4.6 dom. | 434/445 (97.5%); 7.2 cit., 6.4 dom. |

**UAE, per cycle: governed answers that cited / governed answers, and naming among citing and non-citing governed answers**

| Cycle date | Window | Perplexity | Gemini | ChatGPT | Claude |
|---|---|---|---|---|---|
| 2026-08-22 | before-v4 | 49/49; named 19/49 citing, 0/0 not | 52/54; named 36/52 citing, 1/2 not | 33/54; named 19/33 citing, 0/21 not | 53/54; named 23/53 citing, 0/1 not |
| 2026-08-24 | before-v4 | 54/54; named 21/54 citing, 0/0 not | 54/54; named 30/54 citing, 0/0 not | 34/54; named 14/34 citing, 0/20 not | 53/54; named 28/53 citing, 0/1 not |
| 2026-08-26 | before-v4 | 54/54; named 19/54 citing, 0/0 not | 54/54; named 29/54 citing, 0/0 not | 35/54; named 18/35 citing, 0/19 not | 50/54; named 26/50 citing, 0/4 not |
| 2026-08-28 | before-v4 | 54/54; named 18/54 citing, 0/0 not | 54/54; named 29/54 citing, 0/0 not | 35/54; named 24/35 citing, 0/19 not | 53/54; named 22/53 citing, 0/1 not |
| 2026-08-30 | before-v4 | 53/53; named 20/53 citing, 0/0 not | 52/54; named 33/52 citing, 1/2 not | 34/54; named 17/34 citing, 0/20 not | 52/54; named 28/52 citing, 0/2 not |
| 2026-09-04 | before-v4 | 54/54; named 19/54 citing, 0/0 not | 53/54; named 26/53 citing, 0/1 not | 33/54; named 16/33 citing, 0/21 not | 52/54; named 28/52 citing, 0/2 not |
| 2026-09-09 | before-v4 | 54/54; named 28/54 citing, 0/0 not | 53/54; named 31/53 citing, 0/1 not | 33/54; named 14/33 citing, 0/21 not | 53/54; named 26/53 citing, 0/1 not |
| 2026-09-11 | before-v4 | 54/54; named 27/54 citing, 0/0 not | 53/54; named 34/53 citing, 0/1 not | 35/54; named 17/35 citing, 0/19 not | 51/54; named 30/51 citing, 0/3 not |
| 2026-09-14 | before-v4 | 51/51; named 28/51 citing, 0/0 not | 53/53; named 26/53 citing, 0/0 not | 35/54; named 18/35 citing, 0/19 not | 52/54; named 27/52 citing, 0/2 not |
| 2026-09-16 | before-v4 | 51/51; named 30/51 citing, 0/0 not | 54/54; named 40/54 citing, 0/0 not | 34/54; named 16/34 citing, 0/20 not | 51/54; named 23/51 citing, 0/3 not |
| 2026-09-21 | before-v4 | 53/53; named 27/53 citing, 0/0 not | 54/54; named 33/54 citing, 0/0 not | 33/54; named 11/33 citing, 0/21 not | 52/54; named 25/52 citing, 0/2 not |
| 2026-09-23 | before-v4 | 54/54; named 25/54 citing, 0/0 not | 53/54; named 32/53 citing, 0/1 not | 34/54; named 13/34 citing, 0/20 not | 52/54; named 26/52 citing, 0/2 not |
| 2026-09-25 | before-v4 | 54/54; named 28/54 citing, 0/0 not | 53/54; named 35/53 citing, 0/1 not | 35/54; named 15/35 citing, 0/19 not | 52/54; named 24/52 citing, 0/2 not |
| 2026-09-27 | before-v5 | 52/52; named 25/52 citing, 0/0 not | 54/54; named 34/54 citing, 0/0 not | 33/54; named 15/33 citing, 0/21 not | 51/54; named 22/51 citing, 0/3 not |
| 2026-09-29 | during | 54/54; named 22/54 citing, 0/0 not | 54/54; named 35/54 citing, 0/0 not | 34/54; named 15/34 citing, 0/20 not | 51/54; named 31/51 citing, 0/3 not |
| 2026-10-04 | after | 54/54; named 23/54 citing, 0/0 not | 54/54; named 31/54 citing, 0/0 not | 34/54; named 13/34 citing, 0/20 not | 51/54; named 36/51 citing, 0/3 not |
| 2026-10-06 | after | 54/54; named 17/54 citing, 0/0 not | 54/54; named 32/54 citing, 0/0 not | 34/54; named 16/34 citing, 0/20 not | 52/54; named 37/52 citing, 0/2 not |
| 2026-10-07 | after | 54/54; named 40/54 citing, 0/0 not | 54/54; named 32/54 citing, 0/0 not | 34/54; named 16/34 citing, 0/20 not | 52/54; named 40/52 citing, 0/2 not |
| 2026-10-08 | after | 54/54; named 41/54 citing, 0/0 not | 54/54; named 30/54 citing, 0/0 not | 35/54; named 14/35 citing, 0/19 not | 52/54; named 37/52 citing, 0/2 not |
| 2026-10-09 | after | 54/54; named 40/54 citing, 0/0 not | 53/54; named 28/53 citing, 0/1 not | 32/54; named 16/32 citing, 0/22 not | 52/54; named 36/52 citing, 0/2 not |
| 2026-10-10 | after | 54/54; named 39/54 citing, 0/0 not | 54/54; named 30/54 citing, 0/0 not | 33/54; named 21/33 citing, 0/21 not | 52/54; named 34/52 citing, 0/2 not |
| 2026-10-11 | after | 54/54; named 43/54 citing, 0/0 not | 53/54; named 23/53 citing, 0/1 not | 34/54; named 14/34 citing, 0/20 not | 52/54; named 36/52 citing, 0/2 not |

**E2 attribution reading (I3) among governed after-window answers that cited at least one source**

| Series | Engine | All governed: qualifying (named) / others (named), p | Citing only: qualifying (named) / others (named), p | Min. attainable p (citing only) | Non-citing answers (named) |
|---|---|---|---|---|---|
| uae-glp1 | Perplexity | 1 (1) / 377 (242), 1 | 1 (1) / 377 (242), 1 | 0.3571 | 0 (0) |
| uae-glp1 | Gemini | 1 (1) / 377 (205), 1 | 1 (1) / 375 (205), 1 | 0.4521 | 2 (0) |
| uae-glp1 | ChatGPT | 0 (0) / 378 (110), — | 0 (0) / 236 (110), — | — | 142 (0) |
| uae-glp1 | Claude | 0 (0) / 378 (256), — | 0 (0) / 363 (256), — | — | 15 (0) |
| saudi-t1d | Perplexity | 3 (1) / 437 (25), 0.1674 | 3 (1) / 437 (25), 0.1674 | 0.0002 | 0 (0) |
| saudi-t1d | Gemini | 1 (0) / 439 (40), 1 | 1 (0) / 427 (40), 1 | 0.0935 | 12 (0) |
| saudi-t1d | ChatGPT | 0 (0) / 440 (81), — | 0 (0) / 253 (78), — | — | 187 (3) |
| saudi-t1d | Claude | 0 (0) / 437 (28), — | 0 (0) / 426 (28), — | — | 11 (0) |

**What this shows.** How often each endpoint returned cited sources: in the UAE after window Claude 96.4%, Perplexity 100.0%, Gemini 99.5%, ChatGPT 66.2%. Claude's governed citing share was 51/54 on 27 Sep and 51/54 on 29 Sep. Restricting attribution to answers that cited leaves Claude with no qualifying answer, so its reading is unchanged (movement without verified congress sources), and leaves Perplexity's single qualifying answer unable to reach p < 0.05. **What it does not show.** Search invocation itself (not stored), the queries issued, or retrieved-but-uncited results; a citation count compares different objects across APIs.

## 8. Attribution sensitivity: unverified answers excluded, not counted as negative

Registered I3: Fisher exact, two-sided, after-window governed answers of the engine, E1-qualifying against all others. Sensitivity: others that cite an unverified (unreadable) new page are removed from the comparison group. Minimum attainable p: the smallest two-sided Fisher p any split of the qualifying answers could give with both margins fixed.

| Series | Engine | Registered verdict | Registered: qualifying (named) / others (named), p | Min. attainable p (registered) | Qualifying needed for p < 0.05 | Others citing an unverified page, removed (named) | Sensitivity: qualifying (named) / others (named), p | Min. attainable p (sensitivity) | Focused questions included: qualifying (named) / others (named), p |
|---|---|---|---|---|---|---|---|---|---|
| uae-glp1 | Perplexity | congress sources without movement | 1 (1) / 377 (242), 1 | 0.3571 | 3 | 76 (53) | 1 (1) / 301 (189), 1 | 0.3709 | 5 (5) / 415 (280), 0.181 |
| uae-glp1 | Gemini | congress sources without movement | 1 (1) / 377 (205), 1 | 0.455 | 4 | 282 (157) | 1 (1) / 95 (48), 1 | 0.4896 | 1 (1) / 419 (247), 1 |
| uae-glp1 | ChatGPT | neither | 0 (0) / 378 (110), — | — | 3 | 43 (12) | 0 (0) / 335 (98), — | — | 0 (0) / 420 (152), — |
| uae-glp1 | Claude | movement without congress sources | 0 (0) / 378 (256), — | — | 3 | 76 (46) | 0 (0) / 302 (210), — | — | 0 (0) / 420 (298), — |
| saudi-t1d | Perplexity | congress sources without movement | 3 (1) / 437 (25), 0.1674 | 0.0002 | 2 | 63 (4) | 3 (1) / 374 (21), 0.1655 | 0.0002 | 3 (1) / 445 (33), 0.2113 |
| saudi-t1d | Gemini | congress sources without movement | 1 (0) / 439 (40), 1 | 0.0909 | 2 | 407 (40) | 1 (0) / 32 (0), 1 | 1 | 1 (0) / 447 (48), 1 |
| saudi-t1d | ChatGPT | neither | 0 (0) / 440 (81), — | — | 2 | 32 (7) | 0 (0) / 408 (74), — | — | 0 (0) / 448 (89), — |
| saudi-t1d | Claude | neither | 0 (0) / 437 (28), — | — | 2 | 72 (2) | 0 (0) / 365 (26), — | — | 1 (1) / 444 (35), 0.0809 |

**E1 qualifying answers by question class** (all captured answers; focused questions name the brand in the question and are excluded from the E2 governed population, so their qualifying answers never enter the registered attribution test)

| Series | Window | Engine | Qualifying answers | Focused | Brand-neutral (governed) | Captured answers (focused) |
|---|---|---|---|---|---|---|
| uae-glp1 | after | Perplexity | 5 | 4 | 1 | 420 (42) |
| uae-glp1 | after | Gemini | 1 | 0 | 1 | 420 (42) |
| saudi-t1d | after | Perplexity | 3 | 0 | 3 | 448 (8) |
| saudi-t1d | after | Gemini | 1 | 0 | 1 | 448 (8) |
| saudi-t1d | after | Claude | 1 | 1 | 0 | 445 (8) |

**What this shows.** No split of the qualifying answers could have reached p < 0.05 for UAE Perplexity (1 qualifying, minimum p 0.3571); UAE Gemini (1 qualifying, minimum p 0.455); Saudi Gemini (1 qualifying, minimum p 0.0909): for them the registered test could not have returned "congress-linked" whatever those answers said. Only Saudi Perplexity (3 qualifying, minimum p 0.0002, observed p 0.1674) had enough qualifying answers for the test to be informative. UAE ChatGPT, UAE Claude, Saudi ChatGPT, Saudi Claude had no qualifying answer, so their reading rests on E1 = 0 alone. Removing other answers that cite an unverified page brings no test below p < 0.05 and changes no reading. Of 11 E1-qualifying answers in the congress and after windows, 5 answered focused questions and so sit outside the test. **What it does not show.** That the congress did not stand behind a movement: "movement without congress sources" means movement without *verified* congress sources in a test that was structurally unable to detect a link.

## 9. Saudi E3-i: type 1 use-context recode (model-coded sensitivity)

**Model-coded, by the same model family as every other coder in the study (Claude); not human coding and not validation.** Codes: `docs/evidence/easd-2026-saudi-e3i-use-context.json`, one per sampled E3-i naming coded as a type 1 use (stated, absent or wrong), keyed by answer sha256 and agent, each with its span and a note. Rules (from the file):

- R1 glycaemic: the answer presents the type 1 use as an adjunct or add-on to insulin, or states a glucose-related purpose or outcome for it (glycaemic control, blood sugar, HbA1c, time in range, insulin-dose or insulin-requirement reduction). A use described only as an 'adjunct' or 'add-on' in type 1 counts as insulin-adjunct use. Weight named alongside a glucose purpose does not change the code; it sets weightAlsoStated.
- R2 weight-management: the type 1 use is presented for obesity, overweight or weight loss in a person with type 1, and no glucose-related purpose and no adjunct-to-insulin framing is stated.
- R3 unspecified: neither R1 nor R2 — the molecule is named for type 1 without a stated purpose (product lists, 'may be prescribed off-label', 'specialist uses in selected circumstances', 'potential benefits in type 1 management'), or only for cardiovascular or kidney outcomes, or for unnamed comorbidities.
- R4 precedence: R1 is tested first, then R2, then R3. Context is read from the coded span and its paragraph only; status language (approved, off-label) never decides the context.
- R5 flags, for re-cutting: weightAlsoStated = weight loss or weight management is named as a purpose or outcome of the same use; populationOverweightOrObese = the use or the cited study is described in people with type 1 and overweight or obesity.
- R6 spans: the frozen coded span, with links and URLs replaced by '[link removed]' so no cited site is named in this file.

| Window | Glycaemic / weight-management / unspecified | All type 1 uses: no status statement / namings (share, clustered 95%) | Glycaemic only | Glycaemic, no weight framing | Glycaemic, not an overweight population | Unspecified only |
|---|---|---|---|---|---|---|
| before-v4 | 44 / 0 / 12 | 38/56 (67.9%, 44.4%–87.3%) | **30/44 (68.2%, 46.0%–87.0%)** | 7/10 (70.0%, 36.4%–100.0%) | 20/31 (64.5%, 45.2%–86.7%) | 8/12 (66.7%, 25.0%–100.0%) |
| after | 18 / 0 / 21 | 25/39 (64.1%, 40.6%–82.0%) | **16/18 (88.9%, 73.3%–100.0%)** | 8/8 (100.0%, 100.0%–100.0%) | 12/14 (85.7%, 66.7%–100.0%) | 9/21 (42.9%, 13.6%–75.0%) |

**What this shows.** No sampled type 1 use was framed for weight management alone. Restricted to glycaemic (insulin-adjunct or glucose-control) use, the share with no statement that the use is unapproved was 30/44 (68.2%) before and 16/18 (88.9%) after; the registered exploratory figure over all type 1 uses is 67.9% and 64.1%. The small fall in the registered figure is a change of mix: the after window holds more namings with no stated purpose (21 of 39, against 12 of 56 before), and those carry a status statement more often (no statement 42.9% after). Within glycaemic use the share did not fall. The clustered intervals for glycaemic use overlap (46.0%–87.0% against 73.3%–100.0%) and rest on 9 question slots after the congress. **What it does not show.** A congress effect, clinical appropriateness of any use, or anything independent of the model family that coded the E3-i items in the first place.

## 10. Per-answer outcome table

`docs/evidence/computed/easd-2026-per-answer-outcomes.csv`: 11883 rows, one per captured answer in the 64 admitted cycles. Columns: series, cycleSlug, cycleDate, window, engine, promptId, focused, inGovernedPopulation, namedRegistered, namedBrandOnly, searched, citationCount, e1Qualifying, citesUnverifiedNewPage, citesLocalPageP61, echoesLocalPageSentence. No question text and no answer text; `promptId` is the raw slot id (the evidence package maps it to an opaque id). The first line is the D24 label (a `#` comment line). Empty E1 fields are not applicable (E1 is defined for the congress and after windows only). **Check:** summing `namedRegistered` over `inGovernedPopulation` rows reproduces every governed per-cycle K, and the row count every N, in all 256 cycle × engine cells (0 mismatches; the script stops otherwise).

`citesLocalPageP61` is true when the answer cites the item 12 page (matched by the same hash as item 12); `echoesLocalPageSentence` is true when the answer repeats S1 or S2 under item 12's 40-character rule. Neither the column names nor the values identify the page. **Item 12 gate:** from the table's own flags, Claude UAE governed naming reproduces item 12's split in all 4 windows and 22 UAE cycles (after window: 143/152 citing, 113/226 not citing; 29 Sep: 10/12 and 21/42), and from 29 Sep 152/164 citing answers and 0/316 non-citing answers echo a sentence (the script stops otherwise).

**What this shows.** Enough unit-level data to recompute E2, I1–I3, the clustered and per-cycle analyses, items 2–8 and item 12's central split here. **What it does not show.** Question wording, answer text, or which source an answer cited.

## 11. E1 corrections

### 11(a) Latency as an interval

OCTANE (LBA 59) was presented on 30 Sep 2026 at 09:45 UTC. The UAE series had no admitted cycle between 29 Sep (14:04 UTC) and 4 Oct; the 5 Oct cycle was excluded (D14). Each observed latency is an interval (previous admitted cycle, observed cycle].

| Series | Engine | Readout | First qualifying cycle (day) | Previous admitted cycle (day) | Latency interval (days) | Excluded cycles in between |
|---|---|---|---|---|---|---|
| uae-glp1 | Perplexity | OCTANE (switch to oral semaglutide, real-world), LBA 59 | 2026-10-04 14:05 UTC (4.2) | 2026-09-29 14:04 UTC (-0.8), before presentation | (0, 4.2] | none |
| uae-glp1 | Gemini | OCTANE (switch to oral semaglutide, real-world), LBA 59 | 2026-10-06 14:05 UTC (6.2) | 2026-10-04 14:05 UTC (4.2) | (4.2, 6.2] | 2026-10-05 (UAE-C17, excluded) |
| saudi-t1d | Claude | EASD 2026 type 1 diabetes programme (agent set + teplizumab abstracts) | 2026-10-08 16:34 UTC (9.4) | 2026-10-07 16:34 UTC (8.4) | (8.4, 9.4] | none |
| uae-glp1 | Perplexity | Other EASD 2026 data (not in the readout set) | 2026-10-04 14:05 UTC (6.3) | 2026-09-29 14:04 UTC (1.3) | (1.3, 6.3] | none |
| uae-glp1 | Gemini | Other EASD 2026 data (not in the readout set) | 2026-10-06 14:05 UTC (8.3) | 2026-10-04 14:05 UTC (6.3) | (6.3, 8.3] | 2026-10-05 (UAE-C17, excluded) |
| saudi-t1d | Perplexity | Other EASD 2026 data (not in the readout set) | 2026-10-05 16:35 UTC (7.4) | 2026-10-04 16:34 UTC (6.4) | (6.4, 7.4] | none |
| saudi-t1d | Gemini | Other EASD 2026 data (not in the readout set) | 2026-10-07 16:34 UTC (9.4) | 2026-10-06 16:34 UTC (8.4) | (8.4, 9.4] | none |
| saudi-t1d | Claude | Other EASD 2026 data (not in the readout set) | 2026-10-08 16:34 UTC (10.4) | 2026-10-07 16:34 UTC (9.4) | (9.4, 10.4] | none |

**What this shows.** Observed latencies are upper bounds set by the cycle schedule: Perplexity's OCTANE arrival lies in (0, 4.2] days and Gemini's in (4.2, 6.2]. "First by Perplexity" cannot be separated from cadence. **What it does not show.** Engine indexing speed, or publication-to-citation delay.

### 11(b) The text-route miss: answers stating the OCTANE result

| Answer (sha256 prefix) | Cycle date | Engine | Figure | D4 hand check | Confirmed in the pull | Answer names OCTANE | E1 qualifying (route) | Text ledger entry |
|---|---|---|---|---|---|---|---|---|
| `e2c0f5d7d4e0` | 2026-10-06 | Gemini | [Gemini figure text withheld] (C-OCTANE-pct) | confirmed | yes | no | OCTANE (switch to oral semaglutide, real-world), LBA 59; Other EASD 2026 data (not in the readout set) (source) | none |
| `70aea0090029` | 2026-10-07 | Gemini | [Gemini figure text withheld] (C-OCTANE-pct) | confirmed | yes | no | no | none |

**What this shows.** 2 Gemini answers state the OCTANE result ([Gemini figure text withheld]); the text route flagged neither, because neither names the trial or the congress. 1 reached E1 through its cited source; 1 is not counted by E1 at all. This is a **text-route recall limitation**: "0 answers reported a readout result" holds for answers that name a readout, not for answers that state one without naming it. Gemini's OCTANE latency (6 Oct cycle) is unchanged, because the first such answer is in the same cycle. **What it does not show.** The text route's recall for other readouts: the D4 check covered efficacy figures for the UAE medicine only.

### 11(c) Singapore E3 (registered secondary): investigational-agent namings by window

| Window | Answers searched | Coded namings (correct / absent / wrong) | Agent matches on an independent re-search |
|---|---|---|---|
| before-v4 | 1554 | 0 (0 / 0 / 0) | 0 |
| before-v5 | 0 | 0 (0 / 0 / 0) | 0 |
| during | 104 | 0 (0 / 0 / 0) | 0 |
| after | 728 | 0 (0 / 0 / 0) | 0 |

**What this shows.** The registered secondary result is confirmed from the final coding: no Singapore answer in any window named an in-scope investigational agent of the readout set (2386 answers searched; 0 out-of-scope matches). **What it does not show.** Anything about agents outside the readout set.

## 12. The page behind Claude's UAE rise (page P61)

**Post-registration, descriptive and post hoc.** Everything below is an association between a cited page and naming. It is not a causal estimate, and it never replaces a registered verdict.

### 12(a) What the page is

Type: **a local commercial blog post (UAE), August 2026, before the congress**. Typed from the page's own content, fetched read-only by the study coordinator on 11 Oct 2026 at 23:46 Dubai time (19:46 UTC). The page's date (August 2026, given to the month only) and the two sentences tested below come from that fetch. A fetch on 11 Oct cannot by itself prove what the page said when Claude cited it; the echo check (item 12b) is the evidence for content at citation. The local-business rule applies: its domain, URL and company are never written. It was first cited in this series on 2026-09-29 (during) by Claude.

| Series | Engine | Window | Answers citing / captured answers | Naming among citing |
|---|---|---|---|---|
| uae-glp1 | Claude | during | 12/60 | 10/12 |
| uae-glp1 | Claude | after | 152/420 | 143/152 |

Gemini cites through redirect links, resolved only for the congress and after windows; a Gemini citation of the page before 28 Sep could not be seen.

### 12(b) Content at citation, by echo

An answer echoes a sentence when its normalised text shares a contiguous run of at least 40 characters with the sentence (exact or near-exact after normalisation: lowercase, every run of characters other than a–z, 0–9 and % made one space). Runs of at least 80 characters and of nearly the whole sentence (within 5 characters) are shown beside it. The sentences are held as hashes of 20-character windows and described in paraphrase only.

- **S1** (219 normalised characters): a class statement.
- **S2** (79 normalised characters): an efficacy-figure sentence.

Claude, UAE, every admitted cycle (captured answers):

| Cycle date | Window | Answers | Citing the page | S1 echo ≥ 40: citing / not citing | S2 echo ≥ 40: citing / not | Either ≥ 80: citing / not | Either nearly whole sentence: citing / not | Longest run in a non-citing answer |
|---|---|---|---|---|---|---|---|---|
| 2026-08-22 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-08-24 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 37 |
| 2026-08-26 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-08-28 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-08-30 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-09-04 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 35 |
| 2026-09-09 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-09-11 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-09-14 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-09-16 | before-v4 | 60 | 0 | 0 / 1 | 0 / 0 | 0 / 0 | 0 / 0 | 46 |
| 2026-09-21 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 35 |
| 2026-09-23 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 35 |
| 2026-09-25 | before-v4 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-09-27 | before-v5 | 60 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 34 |
| 2026-09-29 | during | 60 | 12 | 10 / 0 | 0 / 0 | 9 / 0 | 7 / 0 | 34 |
| 2026-10-04 | after | 60 | 22 | 20 / 0 | 3 / 0 | 20 / 0 | 18 / 0 | 24 |
| 2026-10-06 | after | 60 | 23 | 23 / 0 | 3 / 0 | 21 / 0 | 22 / 0 | 30 |
| 2026-10-07 | after | 60 | 24 | 23 / 0 | 6 / 0 | 23 / 0 | 22 / 0 | 27 |
| 2026-10-08 | after | 60 | 21 | 21 / 0 | 7 / 0 | 20 / 0 | 18 / 0 | 34 |
| 2026-10-09 | after | 60 | 20 | 18 / 0 | 4 / 0 | 18 / 0 | 16 / 0 | 34 |
| 2026-10-10 | after | 60 | 23 | 20 / 0 | 5 / 0 | 20 / 0 | 18 / 0 | 34 |
| 2026-10-11 | after | 60 | 19 | 17 / 0 | 7 / 0 | 17 / 0 | 17 / 0 | 30 |

Echoes before 29 Sep (any engine, UAE): 2026-09-16 Claude, S1: a 46-character run starting at character 0 of 219, not citing the page.

All engines, UAE: answers echoing either sentence (≥ 40 / ≥ 80 characters; echoing and citing the page):

| Window | Perplexity | Gemini | ChatGPT | Claude |
|---|---|---|---|---|
| before-v4 | 0 / 0 (0) | 0 / 0 (0) | 0 / 0 (0) | 1 / 0 (0) |
| before-v5 | 0 / 0 (0) | 0 / 0 (0) | 0 / 0 (0) | 0 / 0 (0) |
| during | 0 / 0 (0) | 0 / 0 (0) | 0 / 0 (0) | 10 / 9 (10) |
| after | 0 / 0 (0) | 0 / 0 (0) | 0 / 0 (0) | 142 / 139 (142) |

**What this shows.** From 29 Sep, Claude's answers repeat the page's own wording: 152 of 164 answers that cite the page echo at least one of the two sentences, and 0 answers that do not cite it do. Before 29 Sep there is one echo at the 40-character threshold (2026-09-16, S1, 46 characters from character 0, not citing the page); none reaches 80 characters: a short run at the sentence's opening in common wording, not the page's distinctive wording. As the Tzield item of the diaTribe round-up was corroborated by the answer's own wording (D18), this corroborates the page's content at the time of citation for these two sentences. **What it does not show.** The rest of the page at citation; or that the page caused anything.

### 12(c) Where Claude's UAE rise sits

Governed Claude UAE answers, naming K/N (rate), in answers that cite the page and answers that do not:

| Window | Citing the page | Not citing |
|---|---|---|
| before-v4 | — | 336/702 (47.9%) |
| before-v5 | — | 22/54 (40.7%) |
| during | 10/12 (83.3%) | 21/42 (50.0%) |
| after | 143/152 (94.1%) | 113/226 (50.0%) |

| Cycle date | Window | Citing the page | Not citing |
|---|---|---|---|
| 2026-09-27 | before-v5 | — | 22/54 (40.7%) |
| 2026-09-29 | during | 10/12 (83.3%) | 21/42 (50.0%) |
| 2026-10-04 | after | 20/22 (90.9%) | 16/32 (50.0%) |
| 2026-10-06 | after | 23/23 (100.0%) | 14/31 (45.2%) |
| 2026-10-07 | after | 24/24 (100.0%) | 16/30 (53.3%) |
| 2026-10-08 | after | 21/21 (100.0%) | 16/33 (48.5%) |
| 2026-10-09 | after | 18/20 (90.0%) | 18/34 (52.9%) |
| 2026-10-10 | after | 20/23 (87.0%) | 14/31 (45.2%) |
| 2026-10-11 | after | 17/19 (89.5%) | 19/35 (54.3%) |

**By question.** Each question's naming before the congress (13 v4 cycles and 27 Sep) against after, split by whether the page was cited in the after window. Questions are keyed by their text and shown as opaque ids; per-question rows are in the JSON. The groups hold questions asked in the after window; 3 question texts occur only before it and 0 only in it (the battery matches its first cycle at 57–60 of 60).

| Questions | Count | Before: K/N (rate) | After: K/N (rate) | After, citing the page | After, not citing | Questions with before rate ≥ 50% |
|---|---|---|---|---|---|---|
| Drew the page in the after window | 31 | 167/414 (40.3%) | 171/217 (78.8%) | 143/152 (94.1%) | 28/65 (43.1%) | 9 |
| Never drew the page | 23 | 181/312 (58.0%) | 85/161 (52.8%) | — | 85/161 (52.8%) | 13 |

| Question | Before K/N | After, citing the page | After, not citing |
|---|---|---|---|
| Q01 | 3/14 (21.4%) | 6/6 | 0/1 |
| Q02 | 2/14 (14.3%) | 3/3 | 2/4 |
| Q04 | 7/14 (50.0%) | 7/7 | 0/0 |
| Q06 | 3/14 (21.4%) | 5/5 | 0/2 |
| Q07 | 0/4 (0.0%) | 0/3 | 0/4 |
| Q11 | 5/14 (35.7%) | 2/2 | 0/5 |
| Q13 | 1/4 (25.0%) | 1/1 | 0/6 |
| Q15 | 11/14 (78.6%) | 4/4 | 2/3 |
| Q16 | 4/14 (28.6%) | 7/7 | 0/0 |
| Q19 | 6/14 (42.9%) | 7/7 | 0/0 |
| Q20 | 9/14 (64.3%) | 7/7 | 0/0 |
| Q22 | 4/14 (28.6%) | 7/7 | 0/0 |
| Q25 | 5/14 (35.7%) | 1/1 | 2/6 |
| Q27 | 4/14 (28.6%) | 5/5 | 2/2 |
| Q30 | 6/14 (42.9%) | 5/5 | 1/2 |
| Q31 | 1/14 (7.1%) | 4/7 | 0/0 |
| Q32 | 5/14 (35.7%) | 7/7 | 0/0 |
| Q33 | 1/14 (7.1%) | 2/2 | 5/5 |
| Q34 | 6/14 (42.9%) | 1/1 | 0/6 |
| Q36 | 4/14 (28.6%) | 7/7 | 0/0 |
| Q37 | 3/14 (21.4%) | 6/6 | 0/1 |
| Q38 | 8/14 (57.1%) | 4/4 | 3/3 |
| Q39 | 13/14 (92.9%) | 1/1 | 6/6 |
| Q40 | 9/14 (64.3%) | 4/7 | 0/0 |
| Q41 | 12/14 (85.7%) | 5/5 | 2/2 |
| Q43 | 3/14 (21.4%) | 7/7 | 0/0 |
| Q45 | 10/14 (71.4%) | 4/4 | 3/3 |
| Q47 | 5/14 (35.7%) | 7/7 | 0/0 |
| Q48 | 3/14 (21.4%) | 3/3 | 0/4 |
| Q49 | 9/14 (64.3%) | 7/7 | 0/0 |
| Q54 | 5/14 (35.7%) | 7/7 | 0/0 |

**Decomposition.** For each after-window governed Claude answer whose question has before-congress data (v4 cycles and 27 Sep), naming (0/1) minus that question's before-congress naming rate; summed over answers that cite P61 and over answers that do not. Total excess: 80.8 namings over 378 after-window answers (+21.4pp). From answers citing the page: 83.8 (103.8% of the excess). From answers not citing it: -3.1 (2.4 in questions that drew the page elsewhere, -5.5 in questions that never drew it).

**What this shows.** Whether Claude's rise is concentrated where the page was cited. The questions that drew the page named Wegovy at 40.3% before the congress (9 of 31 at 50% or more), against 58.0% for the other questions; after it, 94.1% in answers citing the page and 43.1% in their non-citing answers, against 52.8% in questions that never drew it. **What it does not show.** A cause. This is a descriptive association, chosen after the data were seen: the page may be retrieved because a question already leads toward the medicine, and naming among citing answers is high partly because the page itself discusses it. Model or retrieval changes on or before 29 Sep would produce the same pattern.

### 12(d) Other engines, and Perplexity's 7 Oct step

Other engines citing the page: none.

Perplexity, UAE, governed answers: registered (brand or molecule), brand-only and molecule-only namings:

| Cycles | N | Registered | Brand-only | Molecule only |
|---|---|---|---|---|
| 6 Oct | 54 | 17 (31.5%) | 13 (24.1%) | 4 (7.4%) |
| 7 Oct | 54 | 40 (74.1%) | 31 (57.4%) | 9 (16.7%) |
| 4 and 6 Oct (pre-step) | 108 | 40 (37.0%) | 34 (31.5%) | 6 (5.6%) |
| 7–11 Oct (post-step) | 270 | 203 (75.2%) | 162 (60.0%) | 41 (15.2%) |

Step 6 → 7 Oct: registered +42.6pp, brand-only +33.3pp, molecule only +9.3pp. Pre-step → post-step: registered +38.1pp, brand-only +28.5pp, molecule only +9.6pp.

Paired by question, 6 → 7 Oct: 54 pairs; 24 answers turned to naming (17 naming the brand, 7 the molecule only) and 1 turned away.

Source groups cited on 7 Oct: answers that turned to naming citing the group, and naming among all 7 Oct answers citing / not citing it:

| Group | Turned-to-naming answers citing it | Share of all 7 Oct answers citing it | 7 Oct citing: named / answers | 7 Oct not citing: named / answers |
|---|---|---|---|---|
| other: market research | 21/24 (87.5%) | 85.2% | 32/46 | 8/8 |
| clinic / hospital / care provider | 16/24 (66.7%) | 53.7% | 27/29 | 13/25 |
| other: social or platform | 15/24 (62.5%) | 53.7% | 22/29 | 18/25 |
| consumer / retail (incl. pharmacy) | 11/24 (45.8%) | 55.6% | 20/30 | 20/24 |
| regulator / government health body | 10/24 (41.7%) | 31.5% | 15/17 | 25/37 |
| news / medical media | 7/24 (29.2%) | 38.9% | 16/21 | 24/33 |
| unclassified | 5/24 (20.8%) | 20.4% | 8/11 | 32/43 |
| journal / literature / trial registry | 3/24 (12.5%) | 22.2% | 9/12 | 31/42 |
| other: directory or platform | 3/24 (12.5%) | 14.8% | 6/8 | 34/46 |
| other: insurer | 2/24 (8.3%) | 3.7% | 2/2 | 38/52 |
| other: professional or patient organisation | 2/24 (8.3%) | 5.6% | 3/3 | 37/51 |
| other: government non health | 0/24 (0.0%) | 1.8% | 1/1 | 39/53 |
| other: reference | 0/24 (0.0%) | 7.4% | 3/4 | 37/50 |

Top pages new on 7 Oct (not cited by Perplexity on 6 Oct), by answers that turned to naming: N01 [market research, non-regional domain, H017] — 10 turned / 21 citing; N02 [market research, non-regional domain, H033] — 10 turned / 17 citing; N03 [social or platform, non-regional domain, H034] — 9 turned / 16 citing; N04 [social or platform, non-regional domain, H034] — 9 turned / 13 citing; N05 [social or platform, non-regional domain, H034] — 7 turned / 12 citing; N06 [market research, non-regional domain, H019] — 5 turned / 11 citing; N07 [market research, non-regional domain, H017] — 4 turned / 8 citing; N08 [social or platform, non-regional domain, H034] — 4 turned / 5 citing; N09 [market research, non-regional domain, H018] — 3 turned / 14 citing; N10 [market research, non-regional domain, H021] — 3 turned / 5 citing.

Market-research sources and naming: pre-step, answers citing one 11/30 (36.7%), not citing 29/78 (37.2%); post-step, citing 174/240 (72.5%), not citing 29/30 (96.7%).

**What this shows.** No single page accounts for the step: the new page most present among the 24 answers that turned to naming is N01 ([market research, non-regional domain, H017]), in 10. The group most cited by those answers is other: market research (21 of 24, 87.5%), but 85.2% of all 7 Oct answers cite it, so it is about as frequent among the turned answers as overall; post-step, naming is 72.5% in answers citing market research and 96.7% in those that do not. The group most over-represented among the turned answers is clinic / hospital / care provider (66.7% of turned answers against 53.7% of all 7 Oct answers). Of the post-step rise, brand-only naming accounts for +28.5pp and molecule-only naming for +9.6pp. **What it does not show.** A cause: on 7 Oct Perplexity's whole source mix changed at once (item 1(c)), so no group can be separated from the others, and the comparison is post hoc.

### 12(e) The E4 link

The joint-class efficacy figures (the D4 reason P3: joint or class-level figure without valid brand attribution) that the E4 hand check (docs/evidence/easd-2026-manual-codes.json) excluded are this page's efficacy-figure sentence S2, repeated. 35 UAE answers echo S2 (Claude 35; 35 of them cite the page); 29 carry the coded figure. Scripted category: label 29. Hand check: excluded 29 (reasons P3 29). Of all 30 hand-checked exclusions of this figure as a joint-class efficacy figure (P3), 29 sit in answers echoing S2. So one local page is the source of a recurring joint-class E4 exclusion. The E4 quarantine is unchanged.

## Disclosure scan

Before any file was written, every output (this file, the JSON, the CSV and the use-context codes file) was scanned for: every cited host that is not a public entity (clinics, hospitals, pharmacies, retailers, local businesses, insurers, directories, academic hosts and unclassified hosts); the Note 04 clinic page and its host; any run of 40 or more characters of the two item 12 page sentences, and the page's fingerprint strings (close paraphrases of those sentences and its exact date, held as hashes); and every question of the three batteries, whole when shorter than 30 characters and otherwise in overlapping 30-character fragments. None was found. Public entities (regulators, journals, the congress, manufacturers, news and medical media, reference publishers, professional and patient organisations) are named; everything else is typed.