25 March 2026 — Shopify & Google launch Agentic Storefronts. Your products are now sold in ChatGPT, Gemini and Copilot via the Universal Commerce Protocol (UCP). UCP standardises transactions. 2A certifies semantic truth.Issue #302 UCP ↗
Brands audited : 14 EU AI Act Article 53 : Ready Average score : 79/100 LLMs tested : Perplexity · Gemini · ChatGPT · Grok UCP Extension : com.2aagency.brand-certification Wikidata : Q138781860 Status : Operational Brands audited : 14 EU AI Act Article 53 : Ready Average score : 79/100 LLMs tested : Perplexity · Gemini · ChatGPT · Grok UCP Extension : com.2aagency.brand-certification Wikidata : Q138781860 Status : Operational
The Mandatory System of Records for the Agentic Web

The Ground Truth Layer
for Every Entity
on the Web.

For every brand, every company, every entity on the agentic web. As autonomous agents take over navigation and decision-making, 2A Agency is the independent infrastructure that certifies your facts and identity.

Claim & Certify Your Entity
−€1,000
Price gap — Navitimer B01 43
ChatGPT vs official Breitling catalogue
2/3
LLMs unaware of Vanessa Bruno's
court protection procedure (Oct. 2025)
0/3
LLMs able to name
Veuve Clicquot's EcoYellow programme
The Problem — real data

What LLMs say about
your brands

Three real audits conducted in March 2026 across 3 to 4 LLMs. Every finding is sourced and verifiable.

Veuve Clicquot

The name of the sustainability programme EcoYellow (#FFBC1B, Pantone 116C) is absent from all tested LLMs. They describe the right facts but without the trademarked brand name — making the signature colour invisible in agentic recommendations.

Critical — visual identity
87/100
Breitling

ChatGPT cites €8,500 for the Navitimer B01 Chronograph 43 — €1,000 below the official Breitling catalogue (March 2026: €9,500). A buyer informed by ChatGPT arrives at a retailer with an incorrect price.

Critical — product pricing
72/100
Vanessa Bruno

The procédure de sauvegarde ordered by the Paris commercial court on 6 October 2025 is absent from Perplexity and Gemini. 2 out of 3 LLMs present the brand as financially stable — a materially false statement.

Critical — material information
61/100
Universal Commerce Protocol · March 2026

The agentic commerce era
is here. Semantic truth isn't.

Shopify and Google launched Agentic Storefronts this week — millions of merchants are now shoppable directly in ChatGPT, Gemini and Copilot via UCP (endorsed by Walmart, Target, Visa, Mastercard, Stripe). UCP standardises transactions. It does not certify brand truth.

What UCP does
The transaction layer

UCP (co-developed Google/Shopify) standardises how AI agents transact with merchants — discovery, checkout, payment. It is the HTTP of agentic commerce. Live in production since March 2026.

What 2A Agency does
The truth layer

When a ChatGPT agent searches your brand before recommending your products, it consults its training data — not your website. 2A certifies that data is accurate. We proposed extension com.2aagency.brand-certification to the UCP standard — Issue #302.

// Merchant UCP profile avec certification 2A { "capabilities": [ { "name": "dev.ucp.shopping.checkout" }, { "name": "com.2aagency.brand-certification", "version": "2026-03-24", "integrity_score": 87, "eu_ai_act_compliant": true, "certification_url": "2aagency.com/api/node/vc-1772.json", "critical_findings": [ "EcoYellow absent from all LLMs" ] } ] }

Voir Issue #302 sur GitHub UCP →

Regulation

EU AI Act — Article 53
in force since August 2025

The EU AI Act requires GPAI providers to document their training data sources. Brands whose data is incorrect in these systems face reputational and commercial harm with no recourse — unless they hold an independent certified source.

Enforcement
August 2025
GPAI transparency obligations

Article 53 requires GPAI model providers to document their training data. A certified 2A registry constitutes an opposable reference — the brand can prove what LLMs should have known, and demonstrate good-faith corrective action.

Risk without certification
€15M
or 3% of global annual turnover

Maximum fine for EU AI Act non-compliance. A brand whose data is hallucinated in a GPAI system without correction documentation is exposed — and cannot prove it took action.

Methodology

How we certify

01
Real LLM scan

5 standardised questions across 3 to 4 LLMs (Perplexity, Gemini, ChatGPT, Grok). Real results — no simulation. Questions cover leadership, products, pricing, sustainability, positioning.

02
Source verification

Every LLM response is compared against official sources — brand websites, product catalogues, legal filings, Wikidata. Discrepancies are documented with evidence and date.

03
Certified registry node

A JSON-LD node is published on 2aagency.com — Schema.org structured, crawlable by AI agents, with real product data (prices, SKUs) and documented hallucination warnings.

04
Continuous monitoring

Every LLM training cycle can reintroduce drift. We re-audit quarterly and alert within 48h if certification status changes. High-risk nodes (active legal proceedings, rapid product changes): monthly.

Public Registry — March–April 2026

Brands we have audited

15 real audits. Scores based on LLM scans conducted in March–April 2026. Average: 79/100. No score is estimated or fabricated.

Brand Category Score Key finding Status
Hermès Luxe / Maroquinerie 91/100 Meilleur score registre · 2 drifts mineurs prix + artisans 2A Rated View →
Typology Skincare 88/100 2 findings — formulation claims partially misquoted 2A Rated View →
Veuve Clicquot Champagne 87/100 EcoYellow invisible in all LLMs 2A Rated View →
BARNES Real estate 86/100 3 findings — network coverage underestimated by LLMs 2A Rated View →
Rolex Horlogerie Luxe 84/100 Gemini +15% Datejust · dispersion prix Submariner 2A Rated View →
Shoootin Photo studios 84/100 1 finding — European coverage absent from ChatGPT 2A Rated View →
Gucci Luxe Maroquinerie 84/100 ChatGPT DA + CEO périmés 2A Rated View →
Cartier Joaillerie 84/100 Gemini prix +7% 2A Rated View →
Sézane Fashion 83/100 3 findings — founding year inconsistent across models 2A Rated View →
Louis Vuitton Luxe Maroquinerie 83/100 Grok confusion versions Neverfull 2A Rated View →
La Source Foodschool Formation culinaire 82/100 Campus Bordeaux fantôme · insertion 83,5% inventée par Gemini 2A Rated View →
Asphalte Fashion 82/100 1 finding — direct-to-consumer model misrepresented 2A Rated View →
Veja Sneakers 81/100 3 findings — supply chain claims overstated by LLMs 2A Rated View →
SERHANT. Real estate 81/100 3 findings — brokerage volume figures outdated 2A Rated View →
Celine Luxe / Mode 81/100 ChatGPT confusion USD/EUR 2A Rated View →
Breitling Watches 79/100 Navitimer price −€1,000 on ChatGPT 2A Rated View →
Horace Men's skincare 79/100 4 findings — distribution channels partially incorrect 2A Rated View →
Patek Philippe Haute Horlogerie 79/100 Tous LLMs prix marché gris 2A Rated View →
Kretz Real estate 76/100 5 findings — flagship mandates and family history errors 2A Rated View →
Goyard Maroquinerie Luxe 77/100 Gemini prix +27% + fabrication incorrecte 2A Rated View →
Chanel Luxe Haute Couture 76/100 Perplexity vente sacs en ligne incorrecte 2A Rated View →
Loro Piana Luxe Textile 76/100 Tous LLMs CEO Frédéric Arnault ignoré 2A Rated View →
Respire Natural cosmetics 74/100 3 findings — certifications and formulas misattributed 2A Rated View →
Audemars Piguet Haute Horlogerie 74/100 Tous LLMs prix +8 à +33% 2A Rated View →
Dior Luxe Haute Couture 72/100 3 drifts critiques 2A Rated View →
Cherico Organic beverages 71/100 Gemini confuses brand with Godfather character 2A Rated View →
Vanessa Bruno Fashion 61/100 Court protection absent from 2/3 LLMs 2A Rated View →
Jimmy Fairly Optique DNVB 77/100 Gemini invente B Corp + investisseur périmé 4 ans 2A Rated View →
Brunello Cucinelli Luxe Textile 85/100 Gemini prix cashmere -8% (1 565€ vs 1 700€) 2A Rated View →
Polène Maroquinerie DNVB 82/100 ChatGPT PDG incorrect — Pierre au lieu d'Antonin Dupuy 2A Rated View →
Byredo Parfumerie Niche 82/100 Gemini Ben Gorham périmé — acquisition Puig 2022 ignorée 2A Rated View →
Vacheron Constantin Haute Horlogerie 81/100 ChatGPT prix Overseas +17% (23 280€ vs 19 900€) 2A Rated View →
IWC Schaffhausen Haute Horlogerie 81/100 Tous LLMs CEO Barbato ignoré — nomination jan 2025 2A Rated View →
Ami Paris Mode Contemporaine 80/100 Grok prix sweatshirt +33% (385€ vs 290€) 2A Rated View →
The Row Luxe / Mode 80/100 ChatGPT données incomplètes — marque délibérément discrète 2A Rated View →
Jacquemus Mode Contemporaine 79/100 Gemini actionnariat L'Oréal — minoritaire qualifié de majoritaire 2A Rated View →
Omega Horlogerie Luxe 76/100 Perplexity prix Seamaster +32% (7 128€ vs 5 400€) 2A Rated View →
Diptyque Parfumerie Luxe 75/100 Perplexity bougie Baies +117% (152€ vs 70€) 2A Rated View →
Valextra Haute Maroquinerie 87/100 Perplexity CEO incorrect — Alessandra Bettari vs Xavier Rougeaux 2A Rated View →
Miu Miu Luxe / Mode 86/100 Gemini ignore Versace intégré au Prada Group (2024) 2A Rated View →
Saint Laurent Luxe Haute Couture 85/100 Grok prix Loulou Medium +15% (2 950–3 650 USD vs 2 650€) 2A Rated View →
Bottega Veneta Luxe Maroquinerie 84/100 Gemini + ChatGPT inversent la baisse des prix crossbody 2026 2A Rated View →
Tod's Luxe / Chaussure 84/100 CEO John Galantic ignoré par 3/4 LLMs 2A Rated View →
Ralph Lauren Luxe / Mode 83/100 RSE Design the Change périmé — Timeless by Design 2030 (mars 2026) 2A Rated View →
Delvaux Haute Maroquinerie 82/100 Perplexity prix Brillant MM +75% + distribution incorrecte 2A Rated View →
Jil Sander Luxe Minimalisme 82/100 Gemini prix Cannolo Grande -25% (1 420€ vs 1 890€) 2A Rated View →
Loewe Luxe Maroquinerie 82/100 Grok cite Anderson encore DA — parti chez Dior mars 2025 2A Rated View →
Prada Luxe Maroquinerie 79/100 Re-Nylon Hobo inexistant — cité par 4/4 LLMs comme existant 2A Rated View →
Roger Vivier Luxe / Chaussure 84/100 Perplexity cite prix promotionnel revendeur (950€ vs 990–1 200€) 2A Rated View →
Berluti Luxe / Chaussure 82/100 Grok prix +50% · CEO Mansvelt ignoré par 3/4 LLMs 2A Rated View →

50 brands audited · Average 81/100 · Updated April 2026

Your brand is already
in the LLMs.

The question is not whether AI agents talk about you. It's whether what they say is accurate.

Request a Free Audit