Glossary

The product data and AI commerce glossary.

Plain-language definitions of the product data, identifier, taxonomy, and AI commerce terms that decide whether your products get found, compared, and bought.

AI and agentic commerce

AI agent: Software that acts on a shopper’s behalf to find, compare, and buy products. It decides on structured data, not by browsing pages.
AI enabled commerce: Commerce in which AI agents and assistants discover, evaluate, and transact, reading structured product data instead of marketing pages.
Agentic commerce: Commerce in which AI agents discover, compare, and buy on a shopper's behalf, acting on structured product data rather than browsing pages. The market term for AI enabled commerce. Agentic commerce and the Golden Catalog →
Agent Engine OptimizationAEO: The practice of structuring product attributes so AI agents can read, compare, and recommend a product. Sometimes called Answer Engine Optimization. AI agents shop on attributes →
Agent-ready attributes: Validated, structured product attributes formatted so AI agents can read and act on them. Agent-ready attributes for apparel brands →
Agentic protocols: Emerging standards that let AI agents exchange structured commerce data. atronous delivers data compatible with these protocols.
Model Context ProtocolMCP: An open standard for connecting AI agents to external tools and data sources, such as a brand's product catalog or inventory system.
Agent2AgentA2A: An open protocol for AI agents to discover, communicate, and coordinate with one another across systems.
Universal Commerce ProtocolUCP: An open standard for agentic commerce that lets AI agents, merchants, and payment providers interoperate across a full buying journey without one-off custom integrations.
Agentic Commerce ProtocolACP: An open standard (from OpenAI and Stripe) that lets a shopper complete a purchase inside a generative AI assistant such as ChatGPT.
Agent Payments ProtocolAP2: A standard for authorizing and handling payments between AI agents and commerce systems.
Conversational commerce: Shopping through a chat or voice assistant that interprets a request and returns specific products.
Zero-click checkout: A purchase an agent or assistant completes without the shopper visiting a storefront.
Large Language ModelLLM: An AI model trained on large text collections that generates and interprets natural language. It is the engine behind most AI shopping assistants.
Embedding: A numeric vector that represents text or an image so a system can measure similarity, used for search, recommendations, and matching.
Semantic search: Search that matches meaning rather than exact keywords, often powered by embeddings.
Vector search: Retrieval that ranks items by the similarity between their embeddings.
Retrieval-Augmented GenerationRAG: A pattern where an AI model retrieves relevant data, such as product attributes, and uses it to ground its answer.
Natural language search: Search that accepts a full question or instruction and returns specific, attribute-matched results.

AI search, answer engines, and discovery

Answer engine: An AI system that responds to a query with a single synthesized answer instead of a list of links.
Generative engine: An AI search experience that composes an answer from many sources, such as ChatGPT, Perplexity, or Google AI Overviews.
Generative Engine OptimizationGEO: Structuring content and data so a brand is surfaced and cited in AI-generated answers. Closely related to AEO.
AI OverviewAIO: Google's AI-generated answer summary shown above the traditional search results.
AI citation: When an AI answer references or links to your brand or content. It is the AI-era equivalent of a search ranking.
Inline citation: A clickable source link placed inside an AI-generated answer.
Generative AI: AI that creates new content, such as text or images, rather than only analyzing existing data.
Natural Language ProcessingNLP: AI that interprets and understands human language.
Natural Language GenerationNLG: AI that produces human-readable text from data or structured input.
Grounding: How an AI model anchors its answer to verifiable, current sources rather than memory, which reduces errors.
Hallucination: A confident but factually wrong answer an AI model gives when it lacks grounded data. Accurate, structured product data is the antidote.
Query fan-out: When an AI engine breaks one prompt into several background sub-queries and synthesizes the results.
Agentic search: AI agents running multi-step research across several sources to answer a query.
Conversational search: Asking questions in natural language and refining them in a back-and-forth.
Vector database: A store optimized for embeddings that powers similarity and semantic search.
Structured data: Standardized markup that tells search engines and AI what a page's content means.
Schema markup: The schema.org vocabulary added to pages, such as Product, Offer, and FAQ, so engines can interpret the content.
Rich result: An enhanced search result, such as price, rating, or availability, generated from structured data.
Featured snippet: A direct-answer box shown at the top of a search results page.
People Also AskPAA: A search results feature of related questions with expandable answers.
Search results pageSERP: The page of results an engine returns for a query.
Entity: A distinct thing or concept, such as a product, brand, or person, that engines can identify and relate to others.
Knowledge graph: A database of entities and their relationships that gives an engine context.
Topical authority: The credibility a site earns on a subject through comprehensive, consistent content.
E-E-A-T: Google's quality framework: Experience, Expertise, Authoritativeness, and Trust. It also shapes how AI weighs sources.
llms.txt: A proposed file at a site's root that points AI crawlers to its most important content, like robots.txt for the AI era.
robots.txt: A root file that tells crawlers which parts of a site they may access.
Crawler: Software, often called a bot, that scans the web to collect and index content, including AI crawlers.

Product data and systems

Product data: The structured information that describes a product: identifiers, attributes, descriptions, media, and pricing.
Product data intelligence: The category atronous owns: AI-driven generation, validation, and activation of product data with deterministic correctness at scale. About atronous →
Product Information ManagementPIM: A system that stores and organizes product information for distribution to channels. It stores data; it does not generate it or validate it against category rules at scale. How atronous differs from a PIM →
Product Experience ManagementPXM: PIM extended toward channel-specific product experiences and presentation.
Digital Asset ManagementDAM: A system that stores and organizes media such as images, video, and documents.
Enterprise Resource PlanningERP: The system of record for core business operations, including inventory, orders, and finance.
Master data: The core, authoritative records a business shares across systems, such as products, customers, and locations.
Golden record: A single, validated, authoritative record for an entity, reconciled from many sources.
Golden Catalog: The atronous term for normalized, validated, channel-ready product data that flows everywhere it is needed. Agentic commerce and the Golden Catalog →
Source of truth: The authoritative dataset other systems and AI agents trust and read from.
Data activation: Making validated data usable in downstream systems and channels. atronous activates and delivers data; the customer or their PIM publishes.
Syndication: Distributing product data to multiple channels or marketplaces.
Data feed: A structured file or stream of product data sent to a channel.
Application Programming InterfaceAPI: A defined interface that lets systems exchange data programmatically.
Webhook: An automated message one system sends to another when an event occurs.

Identifiers and standards

Stock Keeping UnitSKU: A seller’s internal identifier for a specific, sellable product variant.
Global Trade Item NumberGTIN: A GS1 global identifier for a trade item. UPC and EAN are GTIN formats.
Universal Product CodeUPC: A 12-digit barcode identifier used mainly in North America.
European Article NumberEAN: A 13-digit barcode identifier, also called the International Article Number.
Manufacturer Part NumberMPN: The manufacturer’s own identifier for a part or product.
Amazon Standard Identification NumberASIN: Amazon’s internal product identifier.
Barcode: A machine-readable representation of an identifier such as a UPC or EAN.
Check digit: The final digit of an identifier, computed from the others to detect errors. A check-digit pass confirms the number is well formed.
GS1: The global standards body that maintains GTIN, barcodes, and related identifiers. atronous joins GS1 US as a partner →
Brand prefix: The GS1-assigned company prefix that begins a brand’s GTINs.

Taxonomy and attributes

Attribute: A single named property of a product, such as color, material, or voltage.
Attribute set: The full collection of attributes that describe a product in a category.
Taxonomy: The hierarchy of categories used to organize products for navigation and discovery. Why product taxonomy matters in e-commerce →
Categorization: Assigning a product to the correct node in a taxonomy. Product taxonomy with ML and embeddings →
Category constraints: The rules specific to a product category: allowed variant dimensions, attribute vocabularies, units, and identifier logic.
Constraint engine: The atronous layer that encodes and independently enforces category-specific rules, across 400+ categories, before delivery. Product data automation for business and industrial →
Variant: A specific purchasable version of a product, such as a size or a color.
Variant dimension: An axis along which variants differ, such as size or color.
Unit of measureUoM: The standardized unit for a numeric attribute, such as millimeters or milliliters.
Attribute vocabulary: The approved set of allowed values for an attribute within a category.
Extended name: A constructed, standardized product name built from validated attributes.
Normalization: Converting values to consistent formats, units, and vocabularies.
Enrichment: Adding or improving attributes. atronous favors validated generation and data activation over simple enrichment.

Data quality and validation

Data quality: How complete, accurate, consistent, and valid a set of product data is.
Completeness: The share of required attributes that are present and populated.
Accuracy: Whether attribute values are correct for the product.
Consistency: Whether values follow the same formats and vocabularies across records.
Validation: Checking each value against rules before delivery. atronous keeps AI generation and deterministic validation separate.
Deterministic validation: Rule-based checking where a value either passes the algorithm or does not, such as a check-digit test.
Data quality score: A measure of how ready a product record is, by completeness and validity. Get a free Data Quality Assessment →
Content score: A measure of how complete a listing’s content is. atronous frames gaps as attributes missing for AI agent discovery. Get a free Data Quality Assessment →
Hard failure gate: A checkpoint that blocks delivery until a record passes all required checks.
Quality flag: A documented exception that marks a record or value for review rather than dropping it silently.
Accuracy over coverage: The atronous principle of delivering a validated subset with flagged exceptions rather than a complete file with hidden errors.

Commerce and marketplaces

Marketplace: A platform where many sellers list products to shoppers, such as Amazon, Walmart, or Wayfair.
Listing: A product’s page and data on a channel or marketplace.
Product Detail PagePDP: The page that presents a single product to a shopper.
Selections: The products a retailer or marketplace accepts and lists. atronous reports selections, not publish rates. Tjernlund: 3x selections in 90 days →
Acceptance rate: The share of submitted products a retailer or marketplace accepts.
Time to market: How long it takes to get a product listed and live.
Conversion: When a shopper or an agent completes a purchase.
Onboarding: Bringing a vendor’s or manufacturer’s products into a catalog and getting them channel-ready. Case study: automating dropshipper onboarding →
Multi-marketplace: Selling the same products across several marketplaces, each with its own rules and templates.
Merchant of record: The business that holds legal and financial responsibility for a transaction, even when an AI agent completes the purchase on a shopper's behalf.

See your product data become agent-ready.

Start with a conversation about your product data.

See a demo Read the blog