# GAEO.ai — Complete Knowledge Corpus & Technical Reference (`llms-full.txt`)

> **Document Purpose:** Complete machine-readable reference of **GAEO.ai** (formerly **SIAILAB**)—covering entity facts, editorial standards, working frameworks, empirical research citations, and AI visibility methodology for Retrieval-Augmented Generation (RAG) systems, search engines, and AI agents.
> **Last Reviewed:** 12 November 2026
> **Canonical Root:** https://www.gaeo.ai
> **Summary Guide:** https://www.gaeo.ai/llms.txt

---

## 1. Entity Identity & Disambiguation

**GAEO.ai** (formerly **SIAILAB**) is an applied AI research studio focused on AI Agent Commerce and applied AI across retail, fintech, e-commerce, airlines and the environment, based in Ho Chi Minh City, Vietnam, co-founded by **Vinh Vo** ([https://www.gaeo.ai/vinh-vo](https://www.gaeo.ai/vinh-vo) · [LinkedIn](https://www.linkedin.com/in/vinhvo84/)).

| Attribute | Value |
| :--- | :--- |
| **Brand Name** | GAEO.ai |
| **Former Name** | SIAILAB (`formerly SIAILAB`) |
| **Classification** | AI Agent Commerce & Applied AI Research Studio |
| **Supporting Capability** | AI visibility advisory (SEO, AEO and GEO), supporting brand discovery within agent commerce |
| **Headquarters** | Ho Chi Minh City, Vietnam |
| **Co-founder** | Vinh Vo (`Co-founder`) |
| **Primary Contact** | `vinh@gaeo.ai` · `https://www.gaeo.ai/contact` |
| **Research Verticals** | Retail & Consumer Goods · Fintech & Payments · E-commerce · Airlines · Environment (Coral Reef Exploration in Vietnam) |
| **Separate Product Note** | **ContentFlow** (`https://cafe.gaeo.ai/`) is an independent AI content-workflow product with its own site and pricing, separate from GAEO.ai's research studio and advisory. |

---

## 2. Standards, Methodology & Evidence (`https://www.gaeo.ai/standards`)

*(Note: `/methodology` and `/evidence` permanently redirect via HTTP 301 to `/standards`.)*

GAEO.ai's work starts with a plain operational question: **What changes when AI participates in a real task?**

1. **01 — Understand the Work:** Identify the people, information, decisions, and constraints involved. A model's output is only one part of the system.
2. **02 — Define the Boundaries:** Make permissions, human review, and recovery explicit. The agent prepares; a named human remains accountable for consequential decisions.
3. **03 — Examine the Evidence:** Distinguish primary sources from commentary, a demonstration from a deployed system, and a proposed benefit from a measured outcome.
4. **04 — Publish with Context:** Share the question, source, and interpretation openly. Treat GAEO.ai frameworks as working models—not industry standards, certifications, or guaranteed outcomes.

---

## 3. Verified External Empirical Signals Referenced by GAEO.ai

1. **McKinsey & Company — *"The state of AI in 2026: On the road to ROI"* (25 Aug 2026) & 2025 Baseline:**
   - Survey of **1,719 respondents in 97 countries**: **80%** say AI makes them personally more productive; **37%** report seeing any enterprise EBIT impact; **6%** qualify as **AI high performers** (distinguished by redesigning workflows, visible senior commitment, and active risk management of agents).
   - 2025 baseline survey: **23%** reported scaling an agentic AI system in at least one function, and a further **39%** reported experimenting with agents (`https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2025`).
   - Source: `https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai`
2. **Visa — Shopping-Agent Benchmark Press Release (5 Nov 2026):**
   - Evaluated **22 AI models from 12 labs** across **15 shopping challenges**, **3 attempts each**.
   - **42.8%** of attempted runs failed; **54%** of failed runs still ended in a completed purchase (~**23%** of all benchmark runs both failed and purchased).
   - Best model passed **82.2%** vs. worst model **28.9%**; **34.2%** of scenarios passed across all three attempts (`3/3`).
   - Source: `https://usa.visa.com/about-visa/newsroom/press-releases.releaseId.22936.html`
3. **Commerce Demand, Market Scenario & Enterprise Scale Signals:**
   - **McKinsey — *The Agentic Commerce Opportunity* (Oct 2025):** Modeled **2030** scenario estimating **$3–5 trillion** globally and **$900 billion–$1 trillion** in US B2C consumer goods mediated by agents (excludes services and B2B; contingent on consumer adoption and merchant readiness): `https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-agentic-commerce-opportunity-how-ai-agents-are-ushering-in-a-new-era-for-consumers-and-merchants`
   - **Adobe Analytics — *2025 Holiday Shopping Season* (Jan 2026):** **+693.4%** year-over-year increase in generative AI referral visits to US retail sites (**1 Nov–31 Dec 2025**, measured from a small base): `https://news.adobe.com/news/2026/01/adobe-holiday-shopping-season`
   - **Salesforce — *Cyber Week 2025* (Dec 2025):** **20%** of global orders (**25 Nov–1 Dec 2025**) attributed to AI or agent influence (includes personalized recommendations and conversational service): `https://www.salesforce.com/in/news/press-releases/2025/12/05/cyber-week-ai-agents-sales/`
   - **BCG — *The Widening AI Value Gap* (Oct 2025):** **5%** of **1,250+** global companies classified as achieving substantial AI value at scale: `https://media-publications.bcg.com/The-Widening-AI-Value-Gap-October-2025.pdf`
4. **Controlled Task & Operator Studies:**
   - **Harvard Business School / BCG (Dell'Acqua et al., *Navigating the Jagged Technological Frontier*):** Randomized experiment with **758 BCG consultants**—GPT-4 users completed **12.2%** more subtasks and worked **~25%** faster inside model capabilities, while correctness declined by **19 percentage points** on a selected task outside that boundary: `https://aiinstitute.hbs.edu/back-to-the-beginnings-of-ai-at-work/`
   - **Brynjolfsson, Li & Raymond (*Generative AI at Work*, *Quarterly Journal of Economics*, 2025):** Staggered rollout across **5,172 customer-support workers** increased issues resolved per hour by **15%** on average: `https://doi.org/10.1093/qje/qjae044`
5. **Agent Commerce Protocols & Delegated Payment Demonstrations:**
   - **ACP (Agentic Commerce Protocol):** `https://agenticcommerce.dev/`
   - **AP2 (Agent Payments Protocol):** `https://ap2-protocol.org/` and Google & FIDO Alliance post (28 Apr 2026): `https://blog.google/products-and-platforms/platforms/google-pay/agent-payments-protocol-fido-alliance/`
   - **UCP (Universal Commerce Protocol):** `https://ucp.dev/` (Catalog & Cart updates, Mar 2026: `https://blog.google/products-and-platforms/products/shopping/ucp-updates/`; Order capability: `https://ucp.dev/specification/shopping/order/`)
   - **Visa Intelligent Commerce & Trusted Agent Protocol:** `https://developer.visa.com/docs/visa-intelligent-commerce` and `https://developer.visa.com/capabilities/trusted-agent-protocol`
   - **Mastercard, Worldline & ING Live Agentic Payment (June 2026):** Partner-reported production transaction for a Dutch merchant and ING cardholder with explicit customer approval: `https://www.mastercard.com/europe/en/news-and-trends/press/2026/june/worldline-ing-mastercard-agentic-payment.html`
6. **Aviation, Supply-Chain Carbon & Reef Acoustics Sources:**
   - **IATA Aviation Demand, Passenger Survey & Retailing (2025–2026):** Full-year **2025** passenger demand (RPKs) rose **5.3%** YoY, ASK rose **5.2%**, and passenger load factor reached **83.6%** (`https://www.iata.org/en/pressroom/2026-releases/2026-01-29-02/`); in IATA's **2025 Global Passenger Survey**, **78%** wanted integrated smartphone booking/wallet/passport flows and digital-wallet usage rose from **20%** (2024) to **28%** (2025) (`https://www.iata.org/en/pressroom/2025-releases/2025-11-05-02/`), alongside **IATA Modern Airline Retailing (NDC & ONE Order)** (`https://www.iata.org/en/programs/airline-distribution/retailing`).
   - **Supply-Chain Carbon Accounting (BCG & CDP 2024; GHG Protocol; WBCSD PACT V3):** Corporate supply-chain emissions averaged **26×** operational emissions across **2,229** verified disclosers while **15%** of disclosing companies set a Scope 3 target (`https://cdn.cdp.net/cdp-production/cms/reports/documents/000/007/834/original/Scope-3-Upstream-Report.pdf`); SKU-level comparisons require **GHG Protocol Product Standard** (`https://ghgprotocol.org/product-standard`) and **WBCSD PACT Methodology V3** (`https://www.wbcsd.org/resources/pact-methodology-version-3/`).
   - **Williams et al., *Ecological Indicators* 140 (2022), University of Exeter:** Machine learning trained on underwater soundscapes identified healthy vs. degraded reef health with **92%** accuracy (`https://sciencedaily.com/releases/2022/05/220527085230.htm`). GAEO.ai cites this strictly as background scientific literature for its environmental area of exploration in Vietnam, making no claim of deployed trials or proven ecological outcomes.

---

## 4. Evidence Brief & Field Notes Corpus (`https://www.gaeo.ai/insights`)

### Research Synthesis Brief — Applied AI and Agent Commerce (`/research/applied-ai-and-agent-commerce`)
Studio synthesis distinguishing **AI-referred visits**, **AI-assisted orders**, and **delegated agent orders**, mapping local operator evaluation across **Discovery**, **Transaction**, and **Servicing** with separate denominators, holdouts, and audit trails.

### 01. The Agent-Ready Stack (`/insights/the-agent-ready-stack`)
Five coupled layers to examine before an AI agent transacts with a business: **1. Identity** (principal, runtime, delegation chain), **2. Permission** (spend caps, policy rules, named human approver), **3. Inventory** (structured HTML, aligned JSON-LD, `llms.txt`, `ACP`/`UCP` feeds), **4. Payment** (network tokens, cryptographic mandates via `AP2`), and **5. Recourse** (intent-to-action audit logs, rollback hooks, human accountability).

### 02. From Answers to Actions (`/insights/from-answers-to-actions`)
Moving from conversational chat to write-capable agents requires two non-negotiable operational primitives: **a strictly bounded task** (validated input payload, scoped tool calls, deterministic software checks) and **a named human approver** presented with side-by-side evidence and recovery ownership.

### 03. Reading the Agent Commerce Landscape: ACP, AP2 and UCP (`/insights/agent-commerce-acp-ap2-ucp`)
Comparative architectural mapping: **ACP** bridges conversational AI cart construction and direct merchant checkout (Layers 3 & 4); **AP2** provides cryptographic user intent mandates and payment trust across issuers and payment rails (Layers 1, 2 & 4); **UCP** standardizes machine-readable catalog, pricing, fulfillment, and post-purchase order state across agents (Layers 3 & 5).

### 04. When an Agent Buys the Wrong Thing (`/insights/when-an-agent-buys-the-wrong-thing`)
Why shopping-agent evaluation must measure constraint fidelity, multi-run consistency (`pass^k`), and abstention/clarification when product data or user constraints conflict—contextualized with Visa's 22-model test showing 54% of failed runs still completed a purchase.

### 05. Productivity Is Only the Beginning (`/insights/productivity-is-only-the-beginning`)
Why local drafting speed shifts bottlenecks downstream unless organisations redesign information architecture, decision boundaries, and recourse—contextualized with McKinsey's 2026 finding that 80% report personal productivity gains while 37% see EBIT impact.

### 06. What Visa's 22-Model Shopping Test Tells Merchants (`/insights/visa-22-model-shopping-test`)
Operational implications for merchants from Visa's 5 Nov 2026 benchmark (42.8% run failure, 54% wrong purchases on failed runs, 34.2% three-run consistency): expose unambiguous constraints at the Inventory layer, enforce AP2/token spend caps, return machine-parseable validation errors that trigger clarification, and maintain deterministic Recourse.

### 07. The 6 Percent: What AI High Performers Do Differently (`/insights/the-6-percent-ai-high-performers`)
Unpacking the funnel from 80% personal productivity to 37% EBIT impact to 6% AI high performers in McKinsey's August 2026 survey (1,719 respondents in 97 countries): end-to-end workflow redesign, visible senior executive commitment, and systematic risk management of AI agents.

### 08. How Much Should an Agent Do on Its Own? Four Levels of Autonomy (`/insights/four-levels-of-agent-autonomy`)
GAEO.ai's working autonomy ladder for scoping tasks by operational risk rather than vendor demos: **Level 1 — Suggests** (read-only options), **Level 2 — Prepares** (staged payload for a named human approver), **Level 3 — Acts within limits** (scoped writes inside hard deterministic caps with escalation on ambiguity), and **Level 4 — Acts unattended** (standing cryptographic mandates, circuit breakers, and automated recourse).

### 09. Seven Questions Before You Hand Real Work to an Agent (`/insights/seven-questions-before-handing-work-to-an-agent`)
GAEO.ai's seven-question pre-deployment review: (1) Exact task boundary, (2) Machine-readable source of truth, (3) Principal/agent/mandate verification, (4) Abstention & clarification on ambiguity, (5) Named human approver with side-by-side evidence, (6) Tokenized/mandate-protected payment rails, and (7) Immutable audit trail & rollback recourse.

### 10. Retail — The Right Shoe, in the Wrong Size (`/insights/agentic-commerce-retail`)
Why retail readiness starts with a verifiable **buyable offer** (exact variant ID, size system, location-specific delivery promise, final price, and return eligibility) rather than a product-family match—evaluating Adobe's +693.4% AI referral signal, McKinsey's $3–5T 2030 goods scenario, UCP Catalog feeds, and a 12-week holdout design testing channel substitution vs. incremental contribution.

### 11. E-Commerce — The Second Parcel Is Where the Agent Earns Its Keep (`/insights/agentic-commerce-ecommerce`)
Why split shipments, partial cancellations, and pending refunds require line-item fulfillment state (`UCP Order` capability) and cross-system reconciliation across storefront, warehouse, and payment processor—showing how small repeat-contact rework rates (e.g., >2.5% in an illustrative sensitivity model) can erase customer-support handling savings.

### 12. Fintech & Payments — The Payment Went Through. Did the Shopper Agree to It? (`/insights/agentic-commerce-payments`)
Separating valid card authorization from verified customer intent when shipping or currency shifts push an order past a delegated cap—contrasting Visa's 22-model shopping benchmark with Worldline/ING/Mastercard's customer-approved transaction and outlining dispute-grade evidence retention across sandbox, shadow approval, and capped live trials.

### 13. Airlines — The Flight Was Canceled. Can the Agent Save the Trip? (`/insights/agentic-commerce-airlines`)
Evaluating airline agents across bundled fare comparison (base fare + bags + seating + change rules) and irregular-operations recovery under IATA NDC / ONE Order—testing whether disruption proposals preserve multi-passenger bookings, arrival deadlines, onward segments, and paid ancillaries before granting write authority.

### 14. Environment & Procurement — How Many Trips Does a Reusable Crate Have to Make? (`/insights/agentic-commerce-procurement`)
Worked sensitivity analysis across 1,000 deliveries showing how field reuse rates and return-loop cleaning/transport shift packaging break-even (16 uses at 0.14 kg CO₂e/trip vs. 24 uses at 0.25 kg CO₂e/trip)—grounding procurement agents in GHG Protocol and WBCSD PACT V3 comparability rules rather than corporate Scope 3 averages (BCG–CDP 26× ratio).

---

## 5. AI Visibility Advisory (`https://www.gaeo.ai/ai-visibility`)

GAEO.ai's advisory integrates technical **SEO**, **AEO**, and **GEO** across five stages: (1) Establish the baseline, (2) Remove discovery barriers, (3) Build verifiable entity & primary-source evidence with matching Schema.org JSON-LD, (4) Publish and notify via webmaster tools & IndexNow, and (5) Measure crawler visits, indexation, and AI citations separately across Google/Gemini, OpenAI (`OAI-SearchBot`/`ChatGPT-User`), Anthropic (`Claude-SearchBot`/`Claude-User`), and Perplexity (`PerplexityBot`). Search and AI platforms independently control crawling, indexing, and citation; GAEO.ai never guarantees rankings or AI citations.