Verification infrastructure for agentic commerce

Product claims should be
inspectable before
AI systems rely on them.

CITAQ is building infrastructure that connects product claims to structured evidence and machine-readable verification records. Its first live module, CRS, helps Shopify merchants identify the product-data gaps that must be resolved first.

CRS for Shopify: available now·Verification Platform: in development
35%
AI recommendations with
unverifiable claims
$41B
Annual consumer fraud
from false product claims
21
Day CRS free trial ·
25 products · no card required
Available now

Why merchants start with readiness.

Before a product claim can be verified, the product itself has to be consistently identifiable, retrievable, and structurally understandable by machine systems. CRS identifies those foundational gaps first.

CRS diagnoses whether product information is machine-ready. The broader CITAQ platform is being built to connect product claims to inspectable evidence and verification records.

Readiness first

Before a product claim can be verified, the product has to be consistently identifiable, retrievable, and structurally understandable. CRS identifies those foundational gaps.

Per-product diagnosis

CRS evaluates each product according to its type — finding what is present, missing, buried in text, or conflicting — then ranks the highest-impact corrections.

Evidence-constrained

CRS never edits your store or invents product facts. It identifies what needs to exist before claims can be linked to evidence and verified.

CRS diagnoses readiness. CITAQ verifies claims.Start 21-Day CRS Trial ->
The structural failure

The model is working correctly.
The source is broken.

Every AI agent — Google AI Mode, ChatGPT, Perplexity — has the same problem: they consume product data written by merchants and reproduce it as fact. With no way to check if the claim is real.

$41B
Global consumer fraud from online shopping annually (2025)
35%
of AI agent product recommendations contain unverifiable claims
$100B+
Projected "hallucination tax" by 2028 — the cost of acting on false claims
0
Products in the world have machine-readable per-claim verification for agents
This is not a model quality problem.The model is functioning correctly when it reproduces merchant copy. Fine-tuning changes nothing. Better prompts change nothing. The failure is at the data source — before the claim reaches the agent's context. The fix is infrastructure that intervenes there.
Independent market signal

The market is waking up to this.

Nate B Jones, a well-known AI strategy creator behind AI News & Strategy Daily with 291K subscribers, described the exact infrastructure gap CITAQ fills.

The Prove-It Economy is Here | And Most Marketers Aren't Ready
AI News & Strategy Daily · Nate B Jones · starts at 4:30
"Agents need you to prove it."
— Nate B Jones, around 4:30 · click to watch
A structured evidence layer for product data that agents can parse
Moving from the attention economy to the interpretation economy
Agents doing the shopping — they need proven, structured data
"Agents don't work like humans. They need you to prove it."
He describes the problem. CITAQ is the infrastructure that solves it.
Three distinct capabilities

Each necessary. None sufficient alone.
The integration is the moat.

01 · STRUCTURED DATA

Machine-readable attributes

Converting unstructured product descriptions, PDFs, and specs into semantically structured formats AI agents can parse without ambiguity. "navy blue" → "#1a237e".

Makes claims readable. Does not verify they are true.
02 · STRUCTURED CLAIMS

Per-claim taxonomy

Extracting and classifying every assertion a product makes — Performance, Safety, Compliance, Origin — and what evidence each type requires to be considered valid.

Identifies what to verify. Does not execute verification.
03 · CLAIMS VERIFICATION

Per-claim verdicts

Evaluating each claim against submitted evidence. Four-way verdict: APPROVE · SUPPRESS · REQUIRE_EVIDENCE · DOWNGRADE. Machine-readable. Cryptographically bound. In development as part of the CITAQ platform.

Platform in development. CRS (readiness layer) is available now.
Platform architecture · in development

10 evaluation stages being built.
Every product claim. Every piece of evidence.

01
Visual Intelligence
Extracts physical attributes from product images. Detects image-based manipulation.
02
Market Signals
Detects claim-level conflicts in real-world reviews. Not star counting — claim-specific analysis.
03
Product Classification
Assigns risk category. Sets evidence thresholds per claim type and product risk level.
04
Claim Evaluation
Per-claim verdict engine with machine-readable reason codes.
05
Jurisdictional Review
Applies regional compliance. A claim legal in the US may be regulated in the EU.
06
Machine-Readability
Scores how accurately AI agents can parse this product. Score ≥75 = AI-ready.
07
Content Enforcement
Applies verdicts to product copy. False claims removed from the canonical record.
08
Identity Compliance
Validates merchant against access controls. Independent of claim quality.
09
Agent Interface
Exposes the evidence vault to AI agents via MCP protocol. Agents verify evidence directly.
10
Public Trust Surface
Renders verified output. Computes tamper-evident content hash. Assigns verification badge state.
Input: unstructured product + evidenceOutput: canonical verified record (CSO) · per-claim verdicts · tamper-evident hash · badge state
Platform architecture · in development

Four verification postures.
Every product. Every claim.

⬤ GREEN

All material claims verified with required evidence. No adversarial signals. Agents can cite with full confidence.

⬤ YELLOW

All verified. Evidence approaching expiry or minor freshness concerns. Agent-trusted with a freshness flag.

⬤ AMBER

One or more performance claims suppressed pending evidence upgrade. Remaining claims still verified.

⬤ RED

Multiple claims suppressed or evidence invalid. Flagged to buyers prominently. Agent citation restricted.

The difference

Same product. Same claim.
Completely different AI answer.

Without CITAQ
Claim: "Waterproof — rated IPX6"
Evidence: none
State: SUPPRESSED
Response: "Claim not independently verified. May not be accurate. Consider checking manufacturer documentation."
With CITAQ
Claim: "Waterproof — rated IPX6"
Evidence: ISO 811 — Intertek Labs
State: APPROVED
Response: "Verified waterproof to IPX6 rating by ISO 811 testing. Certificate on file from Intertek."
Available now · Shopify

CRS for Shopify.
Start with catalog readiness.

CRS is CITAQ's first live module. Connect your Shopify store and get a per-product score showing whether each product is crawlable, identifiable, structurally complete, internally consistent, and supported by the fields expected for its product type.

CRS — Available Now

Citation Readiness Score for Shopify

A 0-100 readiness score with dual CRS-R (retrieval) and CRS-C (citation) scoring. Crawlability check, ranked recommendations, and reverification after edits. No content editing. No copy generation. No store writes.

21-day free trial · 25 products · 50 scans · 1 store · no card required
Commerce standards in motion

Commerce standards are moving toward
more structured, traceable product information.

EU Digital Product Passport
Battery passports mandatory for specified battery classes entering the EU market. Electronics and textiles follow under separate regulation timelines.
Feb 2027
in scope for specified categories
GS1 Digital Link
GS1 describes an industry ambition for retail points of sale to process both linear and 2D barcodes. Linear barcodes will coexist during the transition.
2027 ambition
voluntary guidance
DSA Article 27
EU online marketplaces must provide transparency about recommender systems used on their platforms. Applies to recommender system disclosure.
Active
recommender transparency
EU AI Act
AI systems defined as high-risk under specific product-safety and sensitive use case categories must meet requirements for data quality and documentation.
2026+
phased by risk category
Each standard points toward the same underlying need: product information that is structured, traceable, and machine-readable. That is the infrastructure CITAQ is building.