Kaizen AI Lab · manual test audit · October 7, 2026

U50 is easier to read.
Demand is still the gap.

Reviews and rendered product ratings improved. The catalog is larger, but useful category content, clear brand identity and independently supported claims still need work. The signature hoodie query did not reproduce its August result in this run.

READINESS · WEAK

49/100

August 31: 46 · judgment weighted

GEO READINESS

77/100

23/30 · August 31: 24/30

PRODUCTS WITH REVIEWS

20/53

29 product reviews · prior23

BROWSER HTML RESPONSES

101/101

All200 · command-line results differ

Approval checkpoint. This is the first manual test. Proposed recurring delivery: every other Monday at9: 00 a.m. Eastern, starting October 19, 2026. The automation draft is paused/pending approval. Each run will create a new HTML report, compare the frozen rubric and refresh current research.

Same categories, explicit limits

The PDF’s seven category weights and bands are preserved. It describes a provisional judgment-weighted rubric but does not provide point-level anchors, so this is a consistent editorial reassessment rather than an exact mechanical rescore. The three-point increase is modest; newly available data and evidence can move scores without a site implementation.

Category	Aug 31	Oct 7	Maximum	Reason
Authority and off-page	4	5	25	ACM corroboration and now measurable profile; DR2.5 and little relevant search demand.
Reviews, reputation and proof	6	7	15	More reviews and named partner; coverage remains thin and claims under-evidenced.
Content and competitive substance	6	6	15	Detailed PDPs, thin category pages, empty blog.
Schema and entity integrity	9	10	15	Rendered rating linkage improved; identity unchanged.
Technical crawl and indexation	10	9	12	Browser inventory resolves; client429s, index hygiene and metadata gaps remain.
Agentic and AI readiness	7	7	10	Live agent defaults; verified search-bot inclusion and actual AI answers unmeasured.
Delivery, performance and security	4	5	8	Smaller homepage HTML and more headers; CWV unavailable.
Total	46	49	100	Weak (40–54)

Historical August 3 total: 39 (Poor). Bands: strong85+, good70+, fair55+, weak40+, poor below40. These scores measure observable readiness; they do not estimate rankings, traffic, AI mention probability or market share. Authority and reviews retain40% of the overall weight.

GEO component	Aug 31	Oct 7	Maximum
Discovery metadata	5	4	6
Entity schema	5	5	8
Product/commerce	5	5	5
Rating/review availability	2	3	3
Agent surfaces	4	4	5
AI crawl accessibility	3	2	3
Raw / normalized	24/30 ·80%	23/30 ·76.7%	30

Metadata loses one point for40 missing descriptions; reviews gain one for linked rendered ratings; access loses one for unresolved client-specific429s. Entity, commerce and agent defaults remain level. These are disclosed judgments, not proof that AI visibility fell.

Every baseline finding, retained

Stable IDs carry into future audits. “Partial” includes improvements with remaining work;“not retested” and “unavailable” preserve uncertainty. New findings extend the inventory without changing historic weights.

ID / finding	Status	August 31 evidence	October 7 evidence
F01 · Reviews and rating coverage	partial	23 reviews /16 of47 products; 9 detached rating pages	29 reviews /20 of53 products (37.7% reviewed); 53 Judge.me installs.20 rendered Products contain linked AggregateRating; 0 raw Products do.33 products have no reviews.
F02 · Crawler rate limits	regressed	87/89 first200; all later200; ladder all200	Custom command-line client: 8/102 first200, 94 HTTP429; 16/102200 after retries. Normal Chrome: 101/101 HTML200. Spoofed search/AI UA homepage requests200. Client-specific regression observed; verified provider access remains unknown.
F03 · Collection schema	unchanged	0/21 CollectionPage or ItemList	0/26 in raw HTML. Optional semantic enhancement; absence is not a Google error or proven citation barrier.
F04 · Entity graph IDs	unchanged	No @id across89 URLs	No raw @id across101 HTML pages. Useful for consistent linking when needed, not mandatory on every node.
F05 · Organization identity	unchanged	Domain string name; name, url, sameAs only	Organization name remains u50.com; minimal identity and social links. Use truthful United50/U50 naming, logo and current profiles; no invented founding/address data.
F06 · Shared/absent descriptions	partial	34 shared150-character descriptions	Old broad boilerplate is gone, but40/101 descriptions are absent. Two PDPs share the Forest-All Day Tee description, including foundation-work-out-tee. Prioritize substantive landing pages; utility pages may not need snippets.
F07 · Missing page headings	unchanged	Accessories and Seasonal Looks no H1	Both still lack an H1. Improve visible page labeling; H1 count alone is not a ranking rule.
F08 · Home alias	unchanged	/pages/home200, self-canonical, orphan	Still200 and self-canonical; no incoming source-HTML links within101-page inventory. Resolve intended purpose before redirect.
F09 · Wishlist utility indexing	unchanged	Two indexable orphans, no descriptions	Both200, self-canonical, no noindex or description, zero incoming source links in inventory. Public index eligibility observed; actual Google indexing unknown.
F10 · Copy product URL	partial	Live indexable copy product, orphan	flow-high-waist-body-shorts-copy still200, self-canonical, Product; title HILLCREST ALL WAIST BODY SHORT. Now3 incoming source pages after collection-alias normalization. Verify whether a distinct SKU before consolidating.
F11 · Empty blog	unchanged	Zero article links	/blogs/news still has no article links; 4 main-text words. No substantial editorial publishing observed in inventory.
F12 · Impact claims	partial	No partner named, no dated results or supporting links	Now names American Psychiatric Association Foundation (APAF). Still no U50-specific dated contribution, linked confirmation or substantiation for campaign reach/measurable impact. Bounded APAF search did not verify partnership; this does not disprove it.
F13 · Wholesale inquiry journey	unchanged	Asks for a form that is absent	Fresh browser DOM contains no wholesale inquiry form. Page still asks visitors to complete one; PDF is the remaining route. No forms submitted.
F14 · Homepage response bytes	partial	355KB versus364KB earlier	236, 699 bytes of uncompressed initial HTML this run (~237KB decimal). Improvement indicative; baseline KB convention and request conditions unknown. Does not measure total resources or CWV.
F15 · Backlink/keyword measurement	fixed	Ahrefs unavailable403	Ahrefs now accessible: DR2.5; 1, 014 live backlinks/948 live referring domains; 1 keyword and11 estimated monthly visits. No baseline counts: measurement availability fixed, authority growth not established.
F16 · Core Web Vitals	unavailable	Broker403	Public mobile PageSpeed API429: quota exceeded. No usable LCP/INP/CLS field data or mobile PSI score. API quota is not a U50 performance defect.
F17 · Rating linkage layer	partial	Hidden microdata separate from Product JSON-LD	20 reviewed PDPs now have nested AggregateRating in rendered Product JSON-LD. Initial HTML lacks it; render dependency remains. Mixed syntaxes are not inherently invalid.
F18 · Security headers	partial	nosniff only	Successful browser responses now include HSTS, CSP, X-Frame-Options DENY, nosniff. Referrer-Policy still absent. Error responses differ; adjacent security controls do not imply SEO gains.
F19 · ACM event landing page	unchanged	102 words, description absent, indexable orphan	99 main-text words, description absent, indexable, self-canonical, no incoming source links. An event signup form exists. Official ACM event page corroborates May 16, 2026 workout/pilates event with United50; past event needs lifecycle decision.
F20 · Competitor rating benchmark	regressed	1/5 comparable products has ratings	Current fixed5 URLs: 3/5 raw linked ratings; 4/5 rendered (Tracksmith JS). Only MISSION lacks linked rating. Prior 1/5 is no longer an adequate differentiation claim; baseline rendering method unspecified.
F21 · Agent surfaces	unchanged	llms/agents/UCP/MCP/discovery sitemap live	All inspected surfaces200; UCP advertises2026-08-25; correct /api/ucp/mcp tools/list200. Presence/read-only discovery verified; checkout/provider integration not tested.4/5 competitors llms200, Tracksmith404.
F22 · Thin collection content	unchanged	21 collections, 167–333 total words	26 collections: 141–251 raw total words; 10–119 main-text words. Compare usefulness and buying intent, not a minimum word count; template/count methods differ fromPDF.
F23 · Orphans and internal404	not retested	10 orphans; one internal404 unspecified	12 zero-incoming source-HTML candidates after alias normalization. Not all sitemap/utility orphans require fixing. The PDF does not name its404 URL; limited external-to-inventory sweep was rate limited, so its resolution is not established.
F24 · Duplicate Accessories title	unchanged	Two Accessories titles	Collection and page still share Accessories – u50.com. Align titles with distinct intended pages or consolidate duplication.
F25 · Shipping page label	regressed	Not recorded	New finding: shipping page title SHIPPING but visible H1 RETURNS & EXCHANGES. Review content and navigation consistency.
F26 · Independent authority signal	partial	Only creator/affiliate evidence; no independent publisher surfaced	Official ACM event page names United50 for May 16 event; zero u50.com hrefs on inspected page. This is newly discovered existing evidence, not a newly earned backlink. No independent editorial/Trustpilot/BBB presence surfaced in bounded searches.

Retrieval, authority and actual outcomes

The exact query strings were repeated on the available web-search tool on October 7. These are ordered search retrieval results, not Google rank tracking or verbatim consumer assistant answers. English strings; anonymous tool; locale, underlying model and account unspecified; one run per query. Full outputs are saved. The PDF’s provider/settings are unspecified, limiting direct comparability.

Exact query	Aug 31	Oct 7
best premium heavyweight cotton modal hoodie 430 GSM	U50 position2/8	Absent in30 returned results; absent top8
performance sportswear brand men women recovery travel hoodie	Absent8	Homepage position12/22; absent top8
U50 United50 performance sportswear	Home first +5 U50 pages	Home first +6 other U50 pages in17; 6 U50 pages in top8
best four-way stretch performance tee training to everyday	Exploratory: foundation tee first/8	Absent29; exploratory, excluded from frozen aggregate

Frozen top8 panel: U50 appears for1/3 queries, entirely on the branded query; non-brand category coverage is0/2. The PDF calls its headline “category retrieval1of3” despite separately reporting branded presence. Per-query comparisons are safer than treating those aggregates as identical. No actual ChatGPT, Perplexity, Gemini or Google AI Overview/Mode answer panel was available.

Newly available Ahrefs measurement

Measure	Observed dataset	Interpretation
Domain Rating	2.5/100	Vendor metric, not a Google score
Live profile	1, 014 backlinks ·948 referring domains	All-time2, 696 links /1, 321 domains; no comparable August counts
Organic demand	1 keyword ·11 estimated monthly visits ·$1.94 value	The keyword is “size50”, US position1, size-charts page; last checkedSept12. No observed apparel non-brand keyword in returned dataset.
AI citation corpus	1 citation /1 page across all platforms	ChatGPT, Perplexity, Gemini, Google AI Overview and AI Mode each0 in returned sample. Remaining platform unspecified; not inferred.
Top20 referring domains by DR	18 flagged spam by vendor; all20 have0 dofollow links	Biased sample, not18/20 of entire profile. No inference of purchased links, penalty or need for disavow.
API usage	410 Ahrefs units	Live metrics and estimates, not first-party analytics

Scope: domain/subdomains, global metrics, October 7 snapshot; US keyword row. Ahrefs estimates are neither GA4 visits nor conversion results. One citation in its sampled corpus cannot be combined with the query panel or interpreted as universal AI visibility. Source: local saved connector output (retained locally); no Ahrefs statistics were available in the PDF.

Independent corroboration exists on the official ACM event portal for a May 16, 2026 United50 workout/pilates event. The page has no u50.com link. This supports brand participation, not product quality or link equity. Creator/affiliate and Shop storefront signals remain different evidence classes from independent editorial reviews. Bounded searches did not surface Trustpilot/BBB or independent editorial coverage; absence from search is not proof none exists.

Fixed competitor cohort

Brand / comparable URL	Browser status	Linked rating in raw Product	Linked rating after rendering
Unbound Merino	200	Yes	Yes
MISSION	200	No	No
VIRUS	200	Yes	Yes
Tracksmith	200	No	Yes
Recover Brands	200	Yes	Yes

3/5 raw and4/5 rendered pages have linked ratings, versus the PDF’s1/5. Baseline render method is unknown; do not attribute the whole difference to competitor implementations. All five robots files200; four llms files200, Tracksmith404. Small purposive cohort, not an industry prevalence estimate.

What fresh research changes

Research window: September 7–October 7, 2026; expanded90-day coverage for policy rollouts; older foundations labeled explicitly. The supplied research-seo-geo skill (retained locally) guided evidence classes and the audit. Each statement below separates platform behavior from measured effectiveness.

Current: review AI-assisted content before publishing

Google refreshed its generative content guidance on October 1. Use an editorial checklist for factual product claims, source evidence and original buying guidance. This is a documentation update, not evidence of a new ranking boost. No U50 AI-content use was established.

Measure Google and Bing AI surfaces when access exists

The Google generative AI performance report documents AI Overview/Mode impressions and an August 31 rollout. It may be unavailable at low volume; it does not measure referral conversions. The June 16 Bing visibility preview adds topic/intent and citation-share reporting. These are measurement opportunities; U50 property access is currently unavailable.

Keep raw product data reliable; do not sell schema as citation lift

Google merchant documentation recommends Product data in initial HTML. Bring accurate existing ratings into the initial Product output and keep visible counts consistent. In the May 11 Ahrefs study, 1, 885 treated pages and4, 000 controls showed no clear positive AI Overview citation effect from added JSON-LD across30-day before/after windows. High-citation sample, pretrends and concurrent changes limit transfer to U50; it does not establish causal harm.

Separate search bots, training bots and user retrieval

OpenAI bot documentation and Anthropic crawler documentation distinguish purposes. Add Claude-SearchBot/Claude-User to the measurement plan; the historical ClaudeBot probe covers training. Cloudflare’s July 1 announcement describes new-domain controls effective September 15, including ad-page distinctions. Applicability to existing Shopify U50 is unverified. Resolve access using verified logs, not UA spoofing.

Correct older tactics while retaining baseline measurements

Google AI optimization guidance says llms.txt is not a Google ranking/visibility lever and special GEO schema is unnecessary. Keep agent files as compatibility surfaces, not a moat. The official changelog records FAQ rich results ending May 7, 2026; FAQ content can still help visitors. September 24 VideoObject creator support is relevant only if U50 publishes useful original videos.

Watch experiments; do not adopt promotional claims as proof

A September 25 Ahrefs guide is practical vendor advice, not a controlled study. The December2025 practitioner tests are small and model-specific. A September 23 PR Newswire release promotes its own AEO/GEO offering; no independent U50 effect supports buying it. Prioritize useful category evidence and real reputation, then measure.

Proposed work and acceptance criteria

Owners and dates below are proposals, not assignments or authorized production changes. No site configuration, content or customer workflow was modified.

Priority / change	Proposed owner ·target	Effort / dependency	Definition of done	Success / retest
1 · Repair the wholesale journey and shipping label	Web owner · Oct 12	Low ·existing contact routing	A visible working wholesale form or clear truthful contact path replaces missing-form instructions; shipping H1 matches topic. Submit a controlled test only after implementation authorization.	Form path works; qualified wholesale inquiries tracked; retest browser journey.
2 · Clarify brand identity and substantiate impact	Brand lead +web owner · Oct 19	Low/medium ·approved company facts and partner proof	Organization name United50 with U50 alternate name where supported, current logo/social URLs; publish authorized partner confirmation, dated U50 contribution and sources or revise unsupported quantitative claims.	All assertions traceable; raw identity matches visible About page; retestF05/F12.
3 · Publish truthful raw product ratings and grow review coverage	Web owner +customer experience · Oct 19 markup; Nov 16 coverage	Medium ·Judge.me/theme and truthful review policy	All20 currently reviewed products emit accurate nested AggregateRating in initial Product JSON-LD, matching visible counts; no fake/undisclosed incentivized reviews. Target30/53 reviewed products as an operational proposal, not a ranking threshold.	Raw/rendered counts agree; eligibility validation; review coverage reported with current catalog denominator; retest next audit.
4 · Restore meaningful metadata and index hygiene	Web owner · Oct 19	Medium ·URL-purpose decisions	Unique truthful descriptions on substantive indexable landing pages; review40 omissions and duplicate tee text. Decide home alias, wishlists, event, copy SKU, empty blog; link useful pages and set canonical/redirect/noindex only where appropriate.	No undesired duplicate/utility indexing intent; source/rendered link checks; GSC verification when available.
5 · Build category evidence and earned recognition	Content lead +brand partnerships · Nov 2 pilot; Jan 4 review	High ·product facts, first-party testing, real partner approvals	Publish original material/use-case comparisons on a small category cohort; clarify engineering claims with sources and practical decisions. Prepare legitimate editorial/retail/partner opportunities; no outreach performed in audit.	Qualified non-brand impressions/clicks and conversions; independent references tracked; retest8–12 weeks, seasonality controlled.
6 · Establish outcome and access measurement	Analytics owner +Shopify/web owner · Oct 19	Medium ·authorized GSC/GA4/Bing/Merchant Center/log access	Capture property baselines, non-brand segmentation, AI referrals/conversions, Merchant Center eligibility and verified crawler logs; obtain mobile CWV field data and lab diagnostics. Investigate custom-client429 without assuming consumer bot failure.	Separate impressions/citations/referrals/conversions and lab/field metrics; unknowns remain unavailable, notzero.
7 · Optional semantic and security finishing	Web owner · Nov 2	Low/medium ·template support	Use stable entity IDs where references require them; optional collection semantics matching real content; consider Referrer-Policy and maintain existing headers. Do not score optional syntax as intrinsic defect.	No schema parse errors or mismatched claims; headers verified on successful/error responses; no promised citation lift.

Two bounded experiments

Test	Baseline and treatment	Comparison / window	Success and guardrails
Original category guide	Current26 categories have10–119 main-text words. Add original buying/use-case/material guidance to3 comparable categories; leave3 matched categories unchanged.	8–12 weeks after indexing; match initial demand/product inventory. Use GSC non-brand impressions/clicks and GA4 qualified conversions; requires access.	Compare relative trend with controls, not before/after alone. Track fixed non-brand retrieval as secondary. Confounders: seasonality, stock, promotions, algorithm updates. Revert/repair factual errors or meaningful conversion decline; do not publish filler to hit word counts.
Raw rating delivery	20 reviewed PDPs currently have ratings only in rendered JSON-LD. If rollout supports cohorts, move10 to initial HTML and keep10 unchanged briefly; otherwise staged before/after with explicit limits.	4–8 weeks after recrawl; measure product rich-result eligibility and error changes with authorizedGSC/Merchant Center.	Success: accurate raw parity and no markup errors; AI citations exploratory, not promised. Guardrails: no count inflation, price mismatch or page-speed regression. Confounders: new reviews/product demand. Roll back incorrect markup immediately.

Coverage, gaps and evidence

Site evidence collected October 7, 2026, English public pages. Sitemap index/five children yielded102 records: 101 HTML pages (53 products, 26 collections) plus agents.md. All101 HTML pages were fetched with normal desktop Chrome and examined as initial HTML and rendered DOM. Products allowed1.7 seconds after DOM content loaded; other pages0.6 seconds. Late widgets may change later. Every PDP was examined; review totals use per-product counts, not the homepage/storewide widget.

Product markup: 53/53 raw Product pages; zero JSON-LD parse errors in the inventory. Parsing is not full Google eligibility validation. Canonicals are present; no meta noindex observed. Successful-response X-Robots headers were also captured.
Internal incoming links use source HTML, URL normalization and collection-product alias collapse. The12 candidates are scoped diagnostics; rendered-only or outside-inventory links can change the conclusion. Product/main word counts are reproduced in the CSV, not SEO targets.
Initial custom-client crawl and bounded2/4/8/16 ladder produced429s before skill integration. Retry-After60 was present; short retries did not establish recovery. This test exceeded the low-impact recurring method and will not be repeated. Chrome101/101200 and spoofed UA probes show a client distinction, not proof all crawlers are throttled or allowed.
HTTP and www variants resolve to https://u50.com/ in browser checks. TLS1.3, valid certificate August 28–November 26, 2026; DNS A23.227.38.32. Successful page responses include HSTS/CSP/XFO/nosniff; 429 response headers differ.
Verified robots, llms, agents, UCP version and discovery sitemap; read-only MCP tools/list200. No checkout, cart or transaction calls. Spoofed homepage Googlebot/Bingbot/GPTBot/OAI-SearchBot/PerplexityBot/ClaudeBot requests200, 457–835ms; this one-location timing is not a performance SLA.
Unavailable: GSC, GA4, Shopify admin/logs, Merchant Center, Bing property data, field CWV, usable mobile PSI and authenticated consumer AI answer panel. Mobile experience has not been systematically tested; desktop rendering is not proof of mobile accessibility.
Official Google/OpenAI/Anthropic/Cloudflare/GitHub sources checked; Bing relevant measurement guidance included. No material new U50 action from checked September GitHub or current Anthropic newsroom announcements. Public X exact recent searches yielded no usable dated original; an older original post open failed403. No private bookmarks or exhaustive X scan. Newswire claim deduplicated as issuer promotion.

URL inventory ·searchable101-page appendix

URL	HTTP	Description	H1	Incoming	Reviews	Raw /rendered rating
/	200	Present	1	100	—	0 / 0
/blogs/news	200	Absent	1	0	—	0 / 0
/collections/accessories	200	Absent	1	99	—	0 / 0
/collections/hoodies-sweatshirts	200	Absent	1	100	—	0 / 0
/collections/jackets	200	Absent	1	99	—	0 / 0
/collections/jackets-vests	200	Absent	1	99	—	0 / 0
/collections/joggers-sweatpants	200	Absent	1	99	—	0 / 0
/collections/leggings	200	Absent	1	99	—	0 / 0
/collections/men	200	Present	1	99	—	0 / 0
/collections/mens-swim	200	Absent	1	99	—	0 / 0
/collections/new-arrivals	200	Absent	1	99	—	0 / 0
/collections/new-arrivals-mens	200	Absent	1	99	—	0 / 0
/collections/new-arrivals-womens	200	Absent	1	99	—	0 / 0
/collections/one-pieces	200	Absent	1	99	—	0 / 0
/collections/pants-trousers	200	Absent	1	0	—	0 / 0
/collections/short-sleeve-tops	200	Absent	1	99	—	0 / 0
/collections/shorts	200	Absent	1	99	—	0 / 0
/collections/sports-bra	200	Absent	1	99	—	0 / 0
/collections/united-refine	200	Present	1	1	—	0 / 0
/collections/united-renew	200	Present	1	1	—	0 / 0
/collections/united-reset	200	Present	1	2	—	0 / 0
/collections/united-rise	200	Present	1	2	—	0 / 0
/collections/women	200	Present	1	99	—	0 / 0
/collections/women-hoodies-sweatshirts	200	Absent	1	100	—	0 / 0
/collections/women-joggers-sweatpants	200	Absent	1	99	—	0 / 0
/collections/women-shirts	200	Absent	1	99	—	0 / 0
/collections/women-shorts	200	Absent	1	99	—	0 / 0
/collections/womens-swim	200	Absent	1	99	—	0 / 0
/pages/about	200	Absent	1	99	—	0 / 0
/pages/accessories	200	Absent	0	99	—	0 / 0
/pages/acm-event	200	Absent	1	0	—	0 / 0
/pages/contact	200	Present	1	0	—	0 / 0
/pages/data-sharing-opt-out	200	Present	1	0	—	0 / 0
/pages/design-philosophy	200	Absent	1	0	—	0 / 0
/pages/faqs	200	Absent	1	99	—	0 / 0
/pages/home	200	Absent	1	0	—	0 / 0
/pages/international-orders	200	Absent	1	0	—	0 / 0
/pages/members-club	200	Absent	1	99	—	0 / 0
/pages/our-story	200	Absent	1	99	—	0 / 0
/pages/returns-exchanges	200	Absent	1	99	—	0 / 0
/pages/seasonal-looks	200	Absent	0	99	—	0 / 0
/pages/shipping	200	Absent	1	99	—	0 / 0
/pages/size-charts	200	Absent	1	99	—	0 / 0
/pages/social-responsibility-impact	200	Absent	1	99	—	0 / 0
/pages/swym-share-wishlist	200	Absent	1	0	—	0 / 0
/pages/swym-wishlist	200	Absent	1	0	—	0 / 0
/pages/terms-of-service	200	Absent	1	0	—	0 / 0
/pages/wholesale-inquiries	200	Absent	1	99	—	0 / 0
/products/aligned-fz-offset-hoodie	200	Present	1	2	1	0 / 1
/products/aligned-hv-cropped-cowl	200	Present	1	4	1	0 / 1
/products/aligned-hv-fz-hoodie	200	Present	1	3	1	0 / 1
/products/aligned-hv-fz-mock-nk	200	Present	1	9	0	0 / 0
/products/aligned-hv-pullover-hoodie	200	Present	1	4	3	0 / 1
/products/aligned-hv-straight-leg-sweats	200	Present	1	7	0	0 / 0
/products/aligned-hv-sweat-shorts	200	Present	1	10	0	0 / 0
/products/aligned-puffer-jk	200	Present	1	2	0	0 / 0
/products/breakwater-boardshort	200	Present	1	11	0	0 / 0
/products/coastline-short	200	Present	1	7	0	0 / 0
/products/cropped-polo	200	Absent	1	4	0	0 / 0
/products/every-day-tee	200	Present	1	4	0	0 / 0
/products/first-light-boardshort	200	Present	1	7	0	0 / 0
/products/flow-7-8-rg-waist-leggings	200	Present	1	10	1	0 / 1
/products/flow-crew-pull-over	200	Present	1	5	0	0 / 0
/products/flow-high-waist-body-shorts	200	Present	1	5	0	0 / 0
/products/flow-high-waist-body-shorts-copy	200	Present	1	3	0	0 / 0
/products/flow-high-waist-leggings	200	Present	1	12	0	0 / 0
/products/flow-hoodie	200	Present	1	8	0	0 / 0
/products/flow-no-side-seam-7-8-regular-waist-leggings	200	Present	1	8	0	0 / 0
/products/flow-run-short	200	Present	1	12	1	0 / 1
/products/flow-sleevless-tank-top	200	Present	1	13	0	0 / 0
/products/flow-sports-bra	200	Present	1	13	2	0 / 1
/products/flow-tank-top	200	Present	1	1	0	0 / 0
/products/forest-all-day-tee	200	Present	1	4	0	0 / 0
/products/foundation-7-5-lined-cmft-rs-short	200	Present	1	8	0	0 / 0
/products/foundation-7-5-lined-short	200	Present	1	11	1	0 / 1
/products/foundation-7-5-unlined-cnft-rs-short	200	Present	1	8	1	0 / 1
/products/foundation-7-5-unlined-short	200	Present	1	3	2	0 / 1
/products/foundation-hoodie-tee	200	Present	1	10	0	0 / 0
/products/foundation-mesh-bk-tee	200	Present	1	10	1	0 / 1
/products/foundation-qz-mock-nk	200	Present	1	7	2	0 / 1
/products/foundation-sport-pant	200	Present	1	14	1	0 / 1
/products/foundation-work-out-tee	200	Present	1	13	5	0 / 1
/products/purpose-hv-fleece-pant	200	Present	1	12	1	0 / 1
/products/purpose-hv-fleece-short	200	Present	1	12	1	0 / 1
/products/purpose-hv-full-zip-hoodie	200	Present	1	4	1	0 / 1
/products/purpose-hv-fz-mock-neck	200	Present	1	4	0	0 / 0
/products/purpose-hv-tee	200	Present	1	11	1	0 / 1
/products/purpose-puffy-jk	200	Present	1	2	0	0 / 0
/products/purpose-pullover-hoodie	200	Present	1	10	1	0 / 1
/products/purpose-shell-jk	200	Present	1	1	0	0 / 0
/products/short-body-suit	200	Present	1	5	0	0 / 0
/products/tide-short	200	Present	1	6	0	0 / 0
/products/top-of-the-world-training-tee	200	Present	1	4	0	0 / 0
/products/tote-bag	200	Present	1	1	0	0 / 0
/products/u50-gift-card	200	Present	1	0	0	0 / 0
/products/victoria-bottom	200	Present	1	5	0	0 / 0
/products/victoria-tee	200	Present	1	10	0	0 / 0
/products/victoria-top	200	Present	1	6	1	0 / 1
/products/vista-capri	200	Present	1	10	0	0 / 0
/products/wrap-skirt	200	Present	1	2	0	0 / 0
/products/wrap-top	200	Present	1	3	0	0 / 0

Evidence ledger ·original sources, dates and limitations

ID /claim /source	Publication or update ·event	Surface /outcome	Class /confidence /limits
R01 · Google AI optimization: standard eligibility, useful original content, no special GEO schema; llms not a Google visibility lever official documentation	2026-07-10 2026-06-15 llms clarification	Google Search/AI, global Documented behavior, not uplift	high Eligibility guidance; no guarantee of inclusion.
R02 · Initial HTML Product markup recommended; ratings may be nested in Product official documentation	undated current documentation not specified	Google shopping/search Supported markup	high Parse validity is not a rich-result test or Merchant Center approval.
R03 · GenAI content guidance refreshed; VideoObject creator supported; FAQ rich results discontinued official changelog	2026-10-01 Oct 1 content; Sept 24 video; May 7 FAQ removal	Google Search Documented changes	high Video recommendation conditional on original videos; FAQ removal predates baseline.
R04 · AI content requires useful, accurate editorial review; do not scale low-value pages official documentation	2026-10-01 guidance update Oct 1	Google Search Policy/editorial guidance	high No evidence U50 uses generated content; not a new ranking algorithm.
R05 · Search Console generative AI report measures AI Overview/Mode impressions official help	undated current help 2026-08-31 rollout	Google AI Overview/Mode, global Impressions, not clicks/citations	high Property report may be absent at low volume; actual U50 data unavailable.
R06 · Bing AI visibility preview adds topic/intent, citation share, comparison official announcement	2026-06-16 2026-06-16 preview	Bing AI visibility Observational reporting	high Citation share not rank or traffic; June change newly researched, not new September rollout.
R07 · OAI-SearchBot search, GPTBot training, ChatGPT-User retrieval are separate official documentation	undated current documentation not specified	OpenAI crawlers Crawler controls	high Spoofed UA cannot verify IP/provider access or citations.
R08 · Claude-SearchBot and Claude-User need separate consideration from ClaudeBot official help	2026-04-07 not specified	Anthropic crawler/search/retrieval Documented controls	high Additional identities not directly tested here; logs/IP validation needed.
R09 · Cloudflare new-domain AI traffic options distinguish search, training, agents official changelog	2026-07-01 2026-09-15 policy effective	Cloudflare, new domains/ad pages Access policy	high Do not infer applicability to existing Shopify U50 domain.
R10 · Adding schema showed no clear positive AIO citation effect in vendor matched study original vendor quasi-experiment	2026-05-11 Aug 2025–Mar2026 interventions	Google AI Overviews 1885 treated pages, 4000 controls, 30-day before/after	medium Already high-citation pages; concurrent changes/pretrends; commercial conflict; no causal harm inference.
R11 · Practical AI search optimization guidance refreshed vendor practical guidance	2026-09-25 no separate rollout	Multiple AI search surfaces Advice, not experiment	medium Commercial source; does not establish causal lift.
R12 · Small direct-fetch tests suggest visible HTML matters more than hidden markup original practitioner tests	2025-12-02 2025 tests	5 assistant systems 8 tests	low Old versions, small sample; not universal behavior.
R13 · PR distribution vendor promotes AEO/GEO product issuer press release	2026-09-23 one-year product anniversary	PR distribution Promotional claims	low No independent U50 outcome; not justification for paid PR.
R14 · Official ACM portal lists United50 workout/pilates event independent organization event page	undated 2026-05-16 event	Brand corroboration, US United50 event mentioned; 0 outgoing u50.com links	high Existing event newly discovered; not editorial product review or new link.
R15 · GitHub and Anthropic news checked; no material new U50 SEO action found official changelog	September 2026 entries various	Developer platforms Coverage check	medium Bounded relevant review; also checked https://www.anthropic.com/news; not exhaustive.
R16 · Public X original could not be read public social platform access attempt	2024-05 older post not applicable	X public search Original open403	unavailable Recent exact queries yielded no useful original dated evidence; snippets not used; no private bookmarks access.
R17 · U50 partnership claim names APAF; independent confirmation not found in bounded search first-party page observation	undated not established	U50 impact page Partner named; dated contribution missing	medium First-party claim only; search non-detection does not refute partnership.

All accessed October 7, 2026. Current undated documents are marked undated; source recency is not inferred from copyright or search order. Official behavior supports eligibility/controls, not measured uplift. Vendor observational studies retain conflict and transfer limits.

Reusable delivery package

Metrics JSON · Stable findings · URL inventory CSV · Evidence ledger · Recurring audit specification (retained locally) · Source PDF (retained locally). Raw/rendered snapshots are kept beside these files in the evidence folder.

Next useful step: approve the report format and recurrence, then assign a web owner for the wholesale, identity, rating and metadata fixes. Proposed audit dates: October 19, November 2, November 16, November 30 at9 a.m.Eastern, with daylight saving handled by America/New_York. Every run rechecks all stable findings and refreshes the evidence window; new strategies remain distinct from the frozen rubric.

U50 is easier to read.Demand is still the gap.