> ## Documentation Index
> Fetch the complete documentation index at: https://learn.orbexa.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Why AI agents need a trust verification layer

> The trust dilemma facing AI agents in commerce — why existing trust mechanisms fall short and how OTR solves the verification problem.

# Why AI Agents Need a Trust Layer

## 1.1 The Trust Dilemma for AI Agents

In traditional e-commerce, consumers judge trustworthiness themselves. They evaluate brand reputation, customer reviews, website design, and payment options. These are signals that human intuition handles well.

AI agents have no intuition. They need **machine-readable, quantifiable trust signals**.

Consider a decision an AI agent might face:

```
Consumer: "Buy me those shoes from cheapshoes-deal.xyz"

AI agent's analysis:
  - This domain was registered 3 days ago
  - No SSL certificate
  - No corporate registration on record
  - Privacy policy is a copy-pasted template
  - No return policy

  -> This is very likely a fraudulent site. Should I refuse, or warn the consumer?
```

If an AI agent recommends a fraudulent merchant, the consumer loses money — and loses trust in the AI. For AI platforms, this is an existential threat.

AI agents therefore need an **independent, automated, and tamper-resistant trust assessment system**.

## 1.2 Why Existing Trust Mechanisms Fall Short

| Mechanism                | What It Does                                   | Why It Is Not Enough                                                                                                             |
| ------------------------ | ---------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------- |
| **SSL Certificates**     | Prove that communication is encrypted          | Fraudulent sites can obtain free SSL certificates. SSL only proves "the connection is secure," not "the merchant is trustworthy" |
| **Google Safe Browsing** | Detects malicious websites (phishing, malware) | Only flags "bad" sites — does not evaluate "how good" a site is. A legitimate but low-quality merchant will not be flagged       |
| **BBB / Trustpilot**     | Human reviews and complaints                   | Data is not machine-readable. Reviews can be manipulated. Coverage is limited. AI agents cannot call these services directly     |
| **Domain Age**           | How long the domain has been registered        | Old domains are not necessarily trustworthy. New domains are not necessarily untrustworthy. A single signal is insufficient      |
| **PCI DSS**              | Payment security compliance                    | Only covers the payment layer. Does not assess product quality, policy transparency, or corporate identity                       |

**The core problem**: No existing system evaluates a merchant's overall trustworthiness **across multiple dimensions**, **automatically**, **based on public data**, and delivers results in a **machine-readable format** that AI agents can consume.

This is the problem OTR was built to solve.

## 1.3 OTR's Design Philosophy

OTR (Open Trust Registry) is founded on four principles:

### Principle 1: Publicly Verifiable Data

OTR computes trust scores exclusively from publicly available data sources.

SSL certificates are public. DNS records are public. GLEIF registration data is public. A website's privacy policy is public. Schema.org markup is public.

OTR does **not** use any private data that requires a merchant to self-report (unless the merchant voluntarily authorizes F-dimension verification).

This means anyone can independently verify the basis for any OTR score.

### Principle 2: Multi-Dimensional Assessment

A single dimension is not enough. Good SSL does not mean good policies. Corporate registration does not guarantee data quality.

OTR uses six dimensions: V (Verification), S (Security), G (Governance), T (Transparency), D (Data Quality), and F (Fulfillment).

Each dimension is composed of multiple signals, weighted and aggregated.

### Principle 3: Scores Cannot Be Purchased

This is OTR's most important principle.

Whether or not you buy any ORBEXA product or service has **zero effect** on your trust score. Scores are determined solely by data.

If ORBEXA allowed pay-to-boost scoring, the entire system's foundation of trust would collapse. AI agents will not trust a system where scores can be bought.

### Principle 4: Open Protocol

OTR is an open protocol. The scoring logic is transparent. The API is public. Any AI agent can call it.

Open-source repository: [github.com/yb48666-ctrl/OTR-Protocol-by-orbexa](https://github.com/yb48666-ctrl/OTR-Protocol-by-orbexa)

## Next Chapter

[Chapter 2: OTR Architecture Overview](/en/book-2/ch2-architecture) — Four-layer architecture: Signal Collection, Scoring Engine, ID Generation, Verification API