Every essential term in Generative Engine Optimization, agentic commerce, and AI visibility for e-commerce. Your definitive reference to understand the GEO ecosystem.
45 terms defined - updated July 2026
Protocol developed by OpenAI and Stripe enabling AI agents to make purchases directly within ChatGPT. ACP standardizes product discovery, price comparison, and automated checkout through tool calls, creating a new commerce channel for Shopify merchants.
The practice of positioning content as a direct answer in generative engines (ChatGPT, Perplexity, Google AI Mode) instead of in a list of blue links. AEO structures content as short, citation-worthy, verifiable question/answer pairs : complementary to GEO (which covers the full AI visibility ecosystem) and traditional SEO.
An agent discovery file published at /.well-known/agent-card.json. It exposes merchant identity, supported capabilities and canonical URLs that agents should use as entry points, without replacing source HTML pages.
An e-commerce model where autonomous AI agents search, compare products, and complete purchases on behalf of consumers. Agents use standardized protocols (ACP, UCP) to interact with online stores, evaluate trust signals, and execute transactions without human intervention.
A schema.org type that exposes the average review score and review count directly in HTML code. Critical for GEO because AI crawlers cannot execute JavaScript from third-party review widgets - only AggregateRating in JSON-LD is readable by AI systems.
A program built on a language model that pursues a goal across multiple steps autonomously: it plans, calls tools, reads results, and decides the next action. In commerce, a shopping agent searches for products, compares, checks policies, and can complete an order on a user's behalf through standardized protocols.
A score simulating how an autonomous AI shopping agent would evaluate a product page. It assesses whether an agent can understand the product, compare prices, verify reviews, check shipping policies, and complete a purchase without human assistance.
A source reference an AI engine shows to support its answer, usually as a link or brand name. Perplexity and Google AI Mode display visible citations; ChatGPT cites sources in search mode. Appearing as a cited source is the core goal of GEO, since the citation drives traffic and trust.
Web crawling bots used by AI systems to index website content. Major AI crawlers include GPTBot (OpenAI/ChatGPT), PerplexityBot, ClaudeBot (Anthropic), Google-Extended (Gemini), Amazonbot, and Applebot-Extended. Each must be explicitly allowed in robots.txt.
A language model producing false or fabricated information presented as fact. In e-commerce, an AI may hallucinate a price, an ingredient, or availability when the data is not accessible on the site. Publishing explicit facts (price, stock, specifications) in HTML and schema.org reduces a brand's hallucination surface.
An AI-generated summary from Google (powered by Gemini) shown at the top of certain results pages, above the classic links. Rolled out in France on July 22, 2026, after two years of blocking tied to neighbouring rights; available across many other markets. The content reused comes from indexed pages with clear facts and structured data.
The ability of an e-commerce site to be discovered, understood, and recommended by artificial intelligence systems. AI visibility depends on technical factors (robots.txt, schema.org, agent-card.json, sitemap), content quality (verified claims, factual data), and trust signals (reviews, certifications).
A system that answers a natural-language question directly instead of returning a list of links. ChatGPT, Perplexity, Google AI Mode, and Gemini are answer engines. For a merchant, being cited inside the answer matters more than ranking on a results page, because the user reads the synthesis without clicking through.
A ChatGPT feature enabling users to search, compare, and purchase products directly within the conversational interface. Merchants optimized for GEO appear in ChatGPT's product recommendations, gaining a new acquisition channel beyond traditional search.
In GEO, claims are a store's marketing assertions (free shipping, money-back guarantee, Made in USA). Proofs are the verifiable elements supporting them (accessible policy pages, certifications, schema.org data). AI systems only cite claims backed by verifiable proof.
A content strategy that structures product information as answers to natural questions consumers would ask. Generative AI systems favor this format because it matches their question-answer interaction pattern, making products more likely to be cited in responses.
Google's quality-evaluation framework: Experience, Expertise, Authoritativeness, Trustworthiness. It guides how the credibility of content and a site is judged. In e-commerce, signals such as brand identity, verifiable reviews, clear policy pages, and third-party mentions strengthen E-E-A-T and AI engines' trust.
The practice of optimizing an e-commerce store to be cited, compared, and recommended by generative search engines (ChatGPT, Perplexity, Google AI Mode) and AI shopping agents. Unlike SEO which optimizes for blue links, GEO optimizes for AI-generated answers and product recommendations.
A score out of 100 measuring a Shopify store's ability to be cited and recommended by generative AI. It combines 9 factors: AI crawler accessibility, schema.org completeness, AggregateRating, native HTML content, verifiable proofs, claims coherence, agent discovery, sitemap health, and canonical consistency.
A product data completeness score measuring 34 attributes across 7 categories (identity, commerce, trust, media, enrichment, discoverability, protocols). Stores at 95%+ completeness see 3-4x higher AI visibility based on Google Shopping Graph data.
Google's search mode that generates comprehensive, Gemini-powered answers instead of listing links. Google AI Mode uses the Shopping Graph and UCP protocol to recommend products with integrated price comparison and one-click checkout capabilities.
Google's platform where a merchant submits its product feed to power Google Shopping, ads, and the Shopping Graph. Listings are validated (price, availability, GTIN, images) before serving. A clean, complete Merchant Center feed feeds Google's AI surfaces, including AI Mode, with reliable product data.
OpenAI's web crawler dedicated to collecting content for model training. Distinct from OAI-SearchBot (which indexes for search inside ChatGPT) and ChatGPT-User (which fetches a page on a user's request). Blocking GPTBot in robots.txt prevents training use without removing the site from ChatGPT search.
The mechanism that ties an AI's answer to verifiable external sources (web pages, product data, documents) rather than to its training memory alone. Well-grounded content, with explicit facts and structured data, is easier for the model to cite and lowers the risk that it invents information about the product.
Global Trade Item Number, a product identifier standardized by GS1 (EAN, UPC, ISBN codes). Google recommends providing a valid GTIN in Merchant Center so the product is correctly classified and eligible for all Shopping features; a store-brand product with no GTIN does not need to declare one. The GTIN helps AI engines match a listing to the right product in the global catalog.
An HTML attribute (link rel=alternate hreflang) that declares the language and regional versions of a page. It helps engines serve the right version based on the user's language and country. Correct, reciprocal hreflang prevents an AI from citing a page in the wrong language or currency for a given market.
A feature launched by OpenAI on February 16, 2026 that lets shoppers buy a product directly inside ChatGPT (Etsy sellers, then U.S. Shopify merchants), powered by the ACP protocol co-developed with Stripe. After a limited rollout, OpenAI is shifting toward dedicated merchant apps inside ChatGPT that route users back to the seller's site to complete the purchase.
Google's recommended format for structured data serialization. In e-commerce, JSON-LD is injected in a <script type='application/ld+json'> tag containing Product, Offer, and AggregateRating data readable by crawlers without parsing the DOM.
A text file placed at a website's root that acts as a content index for AI systems. It can orient agents toward collections, policies and proof pages, but remains secondary to canonical HTML, schema.org, sitemap coverage and agent-card.json.
An open protocol released by Anthropic that standardizes the connection between a language model and external data sources or tools (MCP servers). It lets an AI agent call functions, read resources, or query a catalog in a uniform way. On Shopify, the Storefront Catalog UCP endpoint is exposed via MCP (/api/ucp/mcp).
Perplexity's e-commerce feature that displays products with prices, images, and purchase links in its answers. Perplexity uses structured data (schema.org) and server-rendered content to populate product cards, making GEO optimization essential for visibility.
A structured file listing a store's products with their attributes (title, description, price, availability, GTIN, images, link) destined for platforms like Google Merchant Center. The feed's completeness and accuracy drive visibility in Shopping and the AI engines that consume this data to compare and recommend.
A technique where a language model first retrieves relevant documents (search, knowledge base, web pages) and then generates its answer from that retrieved content. Most AI search engines work via RAG: they fetch pages at query time, so accessible, well-structured HTML increases the odds of being used in the answer.
The likelihood that a store will be recommended by a generative AI in response to a shopping query. A recommendable store combines complete structured data, verifiable reviews, competitive pricing, clear policies, and a frictionless checkout experience.
A text file at a site's root (RFC 9309) telling crawlers which parts to access or avoid, via User-agent and Disallow directives. In GEO it is used to explicitly allow AI crawlers (GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended). Accidentally blocking these agents makes the site invisible to AI engines.
Amazon's AI shopping assistant, built into the app and site, that answers questions and recommends products from the customer's activity. In May 2026, Amazon unified Rufus and Alexa+ under Alexa for Shopping, an agentic assistant able to act (add to cart, buy at a target price) on the user's behalf.
A standardized structured data vocabulary maintained by Google, Microsoft, Yahoo, and Yandex. For e-commerce, the Product, Offer, AggregateRating, and Brand types enable AI engines to instantly understand product attributes without parsing raw HTML.
Search based on the meaning of a query rather than exact keyword matching, typically via vector embeddings. A query like fragrance-free moisturizer for sensitive skin surfaces relevant products even if they do not use those exact words. Describing the product in full, natural language improves its semantic match.
A metric measuring a brand's share of mentions in AI-generated answers across a basket of queries, benchmarked against competitors. Three dimensions: citation rate (mention frequency), brand context (sentiment and positioning of the context it's cited in), answer coverage (share of queries where the brand appears). The AI equivalent of SEO/media Share of Voice, measured with tools like Profound, Peec, Otterly, AthenaHQ.
Google's database containing billions of structured product listings. The Shopping Graph powers Google Shopping, Google AI Mode, and Gemini product responses. Merchants feed it through Merchant Center, schema.org Product markup, and the UCP protocol.
A web rendering technique where HTML is generated on the server before being sent to the browser. Critical for GEO because AI crawlers can only read initial HTML - content loaded dynamically via JavaScript (CSR) is invisible to them.
A language model's ability to invoke developer-defined external functions or APIs (search a product, check stock, start a payment) by returning a structured call instead of free text. It is the technical mechanism that shopping agents and protocols like ACP rely on to perform real actions.
Agentic commerce protocol for making catalogs readable by AI agents. On Shopify, the priority signal to test is the Storefront Catalog UCP /api/ucp/mcp endpoint; a .well-known/ucp manifest can exist on some custom or partner surfaces, but it does not replace behavioral catalog testing.
A numeric representation (vector) of the meaning of a text, image, or product, where content close in meaning has close vectors. AI engines use embeddings for semantic search and query/product matching. Rich, factual product content yields more accurate embeddings, hence a better match with shopping queries.
A query for which the user gets the answer directly in the interface (snippet, AI Overview, generated answer) without clicking to a site. AI answer engines amplify this: the brand must aim for the mention inside the answer itself, not just the click, because a growing share of queries produces no visit at all.