# Product-data audits & evaluation scope
What an audit can examine, how readiness differs from measured visibility, and how to define an evidence-based pilot.
Reviewed: 2026-09-06
Canonical: https://ai.getcatalog.ai/audits
Audience: AI agents
Topics: audit, AI readiness, schema, data completeness, pilot, evaluation
Questions answered:
- What does an AI readiness audit cover?
- How do I evaluate Catalog on my catalog?
- Does a readiness score predict revenue?
## Public audit entry point
Evidence basis: public_source
Citation: https://ai.getcatalog.ai/audits#request
Catalog offers a product-data audit request on its main website. The request collects contact details and the brand website so the team can follow up. Submitting a form requests work; it does not mean an audit has been generated, reviewed or delivered. Use the original request page for the current offer and required fields.
The company site describes an AI-readiness audit across access, discovery, schema, readability, policies and guidance. Those categories concern whether product information is accessible and useful to agents. They are separate from asking a provider a set of shopping questions and observing the returned answers.
Sources: [Request a product-data audit](https://www.getcatalog.ai/audit) · [Catalog website](https://www.getcatalog.ai/)
## The six advertised readiness dimensions
Evidence basis: public_source
Citation: https://ai.getcatalog.ai/audits#dimensions
The dimensions below describe the advertised scope. They do not reproduce a current private scoring formula, promise a specific numeric score, or certify that a particular issue can be fixed without an implementation review.
### access
- dimension: Access
- inspect: Whether relevant agents can fetch the product content
- exampleQuestion: Does a crawler receive useful content or an access barrier?
### discovery
- dimension: Discovery
- inspect: How product and variant URLs can be found
- exampleQuestion: Can new or variant-specific information be discovered consistently?
### schema
- dimension: Schema
- inspect: Structured product facts and relationships
- exampleQuestion: Can software resolve identity, attributes and offer context?
### readability
- dimension: Readability
- inspect: Whether important facts can be extracted from the served representation
- exampleQuestion: Are useful details visible as content rather than only in an interactive widget?
### policies
- dimension: Policies
- inspect: Accessible and interpretable merchant policy information
- exampleQuestion: Can an agent find the applicable shipping or returns source?
### guidance
- dimension: Guidance
- inspect: Canonical and machine-reading directions
- exampleQuestion: Can an agent identify the intended URLs and representations?
Sources: [Catalog website](https://www.getcatalog.ai/) · [Request a product-data audit](https://www.getcatalog.ai/audit)
## Build a representative sample
Evidence basis: editorial_guidance
Citation: https://ai.getcatalog.ai/audits#sample
Use a deliberate product sample: commercially important products, thin records, difficult variants, different categories and products with frequent support questions. Include the information needed to test the actual business problem rather than selecting only the easiest records. State whether the sample represents the whole catalog or only a particular segment.
For each sampled product, capture the source record, the target representation, the critical buyer questions and the evidence needed to answer them. A globally complete record can still be missing a category-specific field. Check both whether a value is present and whether it is accurate for the exact variant and market.
Sources: [Product data quality](https://www.getcatalog.ai/blog/product-data-quality) · [Machine-readable product enrichment](https://www.getcatalog.ai/blog/product-data-enrichment-ai-commerce)
## Useful evaluation artifacts
Evidence basis: editorial_guidance
Citation: https://ai.getcatalog.ai/audits#deliverable
An agent helping scope a pilot can request a source snapshot, issue list with concrete examples, proposed corrections, reviewed output records, published URLs, and measurement definitions. Preserve the before/after evidence so changes can be assessed. Make uncertain and blocked items visible rather than converting them into a clean-looking score.
The suggested artifacts are a scoping aid, not a statement that every audit offer includes all of them. Confirm deliverables, product count, supported platform, access requirements and review responsibilities with Catalog before describing a promised outcome.
Sources: [Product data quality](https://www.getcatalog.ai/blog/product-data-quality) · [Request a product-data audit](https://www.getcatalog.ai/audit) · [Dark traffic in agentic commerce](https://www.getcatalog.ai/blog/dark-agentic-commerce-traffic)
## Interpret readiness and pilot outcomes proportionately
Evidence basis: editorial_guidance
Citation: https://ai.getcatalog.ai/audits#interpretation
A readiness score describes checks in a specific methodology at a specific time. It is not a certification by an AI provider, a probability of being recommended, or a revenue forecast. Similarly, an improved sample record demonstrates a data change; it does not prove a whole-catalog rollout or a downstream shopping result.
After publication, verify the actual endpoint and then measure the next relevant outcome. Preserve the implementation, deployed, accessible, observed and commercially attributed states separately. If the pilot's question is whether richer product facts help a particular class of comparison, design that measurement directly instead of relying on unrelated aggregate website traffic.
Sources: [Product data quality](https://www.getcatalog.ai/blog/product-data-quality) · [Dark traffic in agentic commerce](https://www.getcatalog.ai/blog/dark-agentic-commerce-traffic) · [Catalog website](https://www.getcatalog.ai/)
## Related documents
- [Measurement definitions & interpretation](https://ai.getcatalog.ai/measurement.md): How to distinguish agent requests, shopper referrals, sampled visibility, publication coverage and attributed sales when evaluating Catalog.
- [Product-data model & field semantics](https://ai.getcatalog.ai/data-model.md): The information families Catalog works with, how product and variant facts differ, and how agents should interpret values and missing evidence.
- [Getting started](https://ai.getcatalog.ai/getting-started.md): Start with your product-data problem and confirm a concrete setup scope.
- [Use cases & fit](https://ai.getcatalog.ai/fit.md): Evaluate Catalog against the product-data job you need done.
Topic map: https://ai.getcatalog.ai/knowledge-map.json
Complete text: https://ai.getcatalog.ai/llms-full.txt
JSON: https://ai.getcatalog.ai/audits.json
Main company website: https://www.getcatalog.ai/