AI search has grown its own vocabulary fast, and much of it is used loosely: vendor terms, industry labels, file names, and product names all mixed together. This glossary defines the terms an ecommerce brand will actually meet, using the vendor docs where a definition exists, and marking the industry labels where no vendor defines them. Where no vendor defines a term, this guide says so, instead of inventing a definition.
Two rules run through the whole list. A file name is not a promise: llms.txt is a proposal, not a standard, and no engine is documented as reading it. An industry label is not a vendor term: GEO and AEO are words people use online, not products any search engine sells.
Where it goes wrong
Glossary mistakes are not typos. They are decisions made on a wrong definition, and they cost real work.
- GEO is treated as a vendor term. It is not. Google says terms like Answer Engine Optimization and Generative Engine Optimization are common online, and that many of the suggested tricks around them are not effective or supported by how Google Search actually works.
- llms.txt is treated as a standard. It is a community proposal at llmstxt.org, not a standard and not a vendor requirement. The proposal defines llms.txt only. It does not define llms-full.txt, which is a companion convention, not part of the proposal.
- Crawling is confused with indexing. Crawling is a client fetching a page. Indexing is a search engine keeping that page in its index. Google says a page must be indexed and eligible to be shown with a snippet to appear in its AI features.
- Blocking is confused with slowing. Blocking shuts the door. Rate limiting slows the client down with a 429 status and a Retry-After header, while still letting it in. A rule meant to slow a crawler that blocks it instead costs you the visits you wanted.
- Robots.txt is treated as a lock. It is a request. RFC 9309 says plainly that its rules are not a form of access authorization. Enforcement happens at the edge, not in the file.
- Assistant names are used interchangeably. Meta AI is Meta's conversational assistant. Muse is a separate name Meta uses for its agent and model family, not another name for Meta AI. Grok is the assistant, and a crawler token that shares part of the name is a different thing.
The steps
Step 1. Check the vendor doc before you use a term in a plan. The marketer looks up the term in the vendor docs listed in Sources before writing it into a brief. If no vendor defines it, say so in the brief.
Step 2. Use the industry labels as labels, not as products. The marketer uses GEO and AEO to describe a goal, such as being cited in an AI answer, and never as the name of a tool, a score, or a package. There is no vendor formula for GEO, so any GEO score is a made-up number.
Step 3. Keep the file names straight. The developer uses llms.txt for the short index file, robots.txt for the crawling rules, and sitemap.xml for the list of pages worth finding. Each file has one job. The llms.txt file never carries allow or disallow rules.
Step 4. Keep the markup terms straight. The developer uses Product for the item, Offer for the price and availability, and BreadcrumbList for the trail that shows where a page sits. Validate every block with the Rich Results Test before it goes live.
Step 5. Use the crawling terms when you talk about load. The developer says crawl-delay for the robots.txt request, rate limiting for the edge rule, and 429 for the status the server returns. Mixing them up is how a slow-down request turns into a block nobody intended.
Step 6. Revisit the glossary when a vendor doc changes. The marketer rechecks the terms against the vendor docs when a vendor updates its guidance. A glossary that is never rechecked becomes a list of stale definitions.
The terms
AEO, answer engine optimization. Industry wording for optimizing so a brand is cited in an AI-generated answer. Google names AEO as a term common online. No vendor sells AEO as a product.
AI Overviews and AI Mode. AI-generated responses in Google Search that answer a question with links to supporting pages.
BreadcrumbList. The schema.org type for the trail of links that shows where a page sits in a site hierarchy.
Crawl-delay and rate limiting. Crawl-delay is a non-standard robots.txt line that asks a crawler to wait a number of seconds between requests. It is not part of RFC 9309, and Google ignores it. Bing documents support for it. Rate limiting is slowing a client down at the edge, usually with a 429 status and a Retry-After header, while still letting it in.
Crawler. An automated client that fetches pages, usually to build an index or to gather content. Search engines and AI systems both run crawlers.
GEO, generative engine optimization. Industry wording for making a page more likely to be surfaced and cited inside an AI-generated answer. No vendor defines GEO, publishes a GEO formula, or sells GEO as a product.
Assistants. ChatGPT is the assistant product from OpenAI. Claude is the assistant product from Anthropic. Gemini is the assistant and model product from Google. Copilot is the assistant product from Microsoft. Perplexity is the answer engine product from Perplexity. Grok is the assistant product from xAI. Meta AI is Meta's conversational assistant, and Muse is a separate name Meta uses for its agent and model family, not another name for Meta AI. Each entry identifies the product only. What each one reads, cites, or fetches is defined by its own vendor docs, not by this guide.
Index. The store of pages a search engine keeps after crawling, from which it serves results.
JSON-LD. A format for writing structured data as a script block in a page, using the schema.org vocabulary. Google recommends JSON-LD where a setup allows it.
llms.txt. A short Markdown file proposed at llmstxt.org that gives an AI reader a map of a site: an H1 with the site name, a blockquote summary, and H2 sections of links. The proposal is a community proposal, not a standard, and no engine is documented as reading it.
llms-full.txt. A single file that holds the full text of key pages in one document. It is a companion convention from documentation platforms, not part of the llms.txt proposal, not a vendor requirement, and not documented as read by any engine.
Product and Offer. Product is the schema.org type for an item a store sells. Offer is the type for its commercial terms: the price, the currency, and the availability. A Product block carries its price and stock inside a nested Offer.
RAG and query fan-out. RAG, retrieval-augmented generation, is a technique that grounds AI responses in pages retrieved from a search index. Query fan-out is a set of concurrent, related queries a model generates to fetch more information for one user question. Google names both as techniques its AI features use.
robots.txt. A plain text file at the root of a site that tells automated clients which paths they may fetch. Its rules are defined by RFC 9309, which says they are not a form of access authorization. Google supports only the user-agent, allow, disallow, and sitemap fields.
Schema.org. The shared vocabulary for structured data, used to label the facts on a page in a machine-readable way. Types such as Product, Offer, and BreadcrumbList come from schema.org.
SEO, search engine optimization. The practice of making pages findable, readable, and citable in search. Google says SEO best practices continue to be relevant for generative AI search.
Sitemap. A machine-readable list of the pages on a site worth finding, with the date each page last changed.
Snippet. The short piece of a page a search engine is allowed to show in results.
Structured data. Machine-readable markup on a page, usually JSON-LD, that labels the facts on the page using the schema.org vocabulary. Google says it is not required for generative AI search.
User-agent. The name a client sends to identify itself, such as a crawler name in an HTTP header and in a robots.txt group. RFC 9309 calls the robots.txt name a product token.
The machine-readable piece
This is the shape of a robots.txt file, following RFC 9309 and the fields Google supports. Replace the placeholders with your own paths.
User-agent: *
Allow: /
Sitemap: https://[yourdomain.com]/sitemap.xml
This is the shape of an llms.txt file, following the proposal at llmstxt.org. Replace the placeholders with your own pages.
# Example Plant Supply
> Example Plant Supply sells indoor plants, pots, and care supplies, shipping across the country.
## Shop
- [Indoor plants](https://[yourdomain.com]/collections/indoor-plants): Indoor plants, listed with light needs and size.
## Guides
- [Watering guide](https://[yourdomain.com]/pages/watering-guide): How often to water, and how to tell when a plant is thirsty.
Check both files after you publish them. Open your domain followed by robots.txt and confirm the rules are the ones you meant. Open your domain followed by llms.txt and confirm the H1 names the brand and every link loads.
What it looks like when it works
A glossary is not a page shoppers see. The change is in the room where the work gets planned: the marketer, the developer, and the contractor all use the same words to mean the same things. A brief that says slow the crawler produces a rate limit, not a block. A pitch that promises a GEO score gets asked for the vendor doc behind the number, and the pitch fails that test.
Say you run a plant and garden store, selling indoor plants, pots, and care supplies, with a watering guide that shoppers ask about often. The marketer hires a contractor who promises GEO, AEO, and a full llms.txt rollout in one package. The marketer opens this glossary and checks each term: GEO has no vendor formula, AEO is an industry label, and llms.txt is a proposal no engine is documented as reading. The marketer rewrites the brief in plain words instead: make the watering guide worth citing, keep the prices accurate, and check that the crawlers can read the store. When the contractor's report arrives, the marketer reads it against the glossary and keeps the checks that trace to a vendor doc, and drops the score that does not. A shopper asks an assistant how often to water a fiddle leaf fig. The assistant can find the current watering guide and cite the store with the steps as written.
The fundamentals
These mechanics are shared across every engine and covered in full in their own guides:
- •The llms.txt format and a complete sample, see our guide to writing a good llms.txt file
- •The markup AI systems read to verify facts, see our product schema guide
- •The complete crawler rule reference, see our robots.txt guide for AI search
What this guide does not cover
This guide covers the terms an ecommerce brand meets in AI search, defined from vendor docs where a definition exists. It does not cover the full technical audit that checks every crawler and every page, the transaction layer that lets an agent complete a purchase, or the per-engine playbooks for getting each assistant to find you. It does not cover how to write any single file or markup type in full. Those are separate jobs, and this guide is the map, not the whole trip.
Check your own store now
Run the free scan. One click, no signup, and it shows whether AI crawlers can read your store today: aizii.ai/aeo-scan
Run the free scanSources
- llms.txt proposal (specific): https://llmstxt.org
- RFC 9309, Robots Exclusion Protocol (specific): https://www.rfc-editor.org/rfc/rfc9309.txt
- Google Search Central, Optimizing your website for generative AI features on Google Search (specific): https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- Google Search Central, How Google interprets the robots.txt specification (specific): https://developers.google.com/search/docs/crawling-indexing/robots/robots_txt
- Google Search Central, Breadcrumb structured data (specific): https://developers.google.com/search/docs/appearance/structured-data/breadcrumb
- Schema.org, BreadcrumbList type (specific): https://schema.org/BreadcrumbList
- Schema.org, Product type (specific): https://schema.org/Product
- Schema.org, Offer type (specific): https://schema.org/Offer
Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.
