AI agent glossary: terms in plain English

The terms, A to Z

ACP (Agentic Commerce Protocol)

An open standard from OpenAI and Stripe for agents to check out with merchants; the basis of ChatGPT's commerce features. More on ACP (Agentic Commerce Protocol).

AEO (answer engine optimization)

Tracking and improving how a brand appears in AI answers. It measures mentions and citations, not whether an agent can finish a task on your site. More on AEO (answer engine optimization).

Agent experience (AX)

How an AI agent experiences your business once it arrives: whether it can find, read, act on, pay and trust your site. The business counterpart of user experience. More on Agent experience (AX).

Agent journey

A recorded run in which an agent is given a real task on a site, with every step, stall and outcome logged. More on Agent journey.

Agent readiness

How well a site supports AI agents across finding, reading, acting and paying. Scanners turn it into a score; the self-check turns it into decisions. More on Agent readiness.

Agent signup

An agent creating an account and getting credentials for a person, with that person's approval. More on Agent signup.

Agentic checkout

A purchase completed inside an AI assistant or by an agent, with the merchant still selling the item. More on Agentic checkout.

AgentReady

An open, MIT-licensed list of what a product should implement so agents can find, read and act, maintained by Ora with partners. It names practices and scores no one. More on AgentReady.

AI agent

Software that carries out a task on the web for one person, right now: it reads pages, fills forms, compares options and sometimes pays. Different from a crawler, which collects pages in bulk. More on AI agent.

AI crawler

A bot that collects pages in bulk on its own schedule, to train a model or build an AI search index. Blocking it changes what an AI system knows about you later, not a customer's request now. More on AI crawler.

AI search crawler

A crawler that indexes pages so an AI search product can answer questions and cite them later. More on AI search crawler.

API

A documented way for software to ask your systems for data or to do something, without using the web pages. More on API.

Bot management

Tools at your CDN, host or firewall that identify automated visitors and allow, challenge or block them. It is where a robots.txt decision is actually enforced. More on Bot management.

Buy for Me

Amazon's feature that completes a purchase of an eligible product on another retailer's site for the customer. More on Buy for Me.

Connector

A link between a personal agent and a service, so the agent can use that service directly with the user's permission. More on Connector.

Content Signals

A line in robots.txt that states how your content may be used, for example for search, AI input or training. More on Content Signals.

Grounded answer

An agent's answer built from the business's own pages rather than from search results or other sites. More on Grounded answer.

HTTP status code

The three-digit code a server returns with every page, such as 200 for found and 403 for forbidden. Agents trust these codes. More on HTTP status code.

Initial HTML

The page as your server first sends it, before any scripts run. Many agents read only this version. More on Initial HTML.

JavaScript rendering

Building page content in the visitor's browser with scripts after the page loads. Readers that do not run scripts see an empty or partial page. More on JavaScript rendering.

llms.txt

A proposed plain-text file that points AI tools to a site's most useful pages, written for language models. More on llms.txt.

Markdown negotiation

A site returning a plain markdown version of a page when an agent asks for one with the header Accept: text/markdown. More on Markdown negotiation.

MCP (Model Context Protocol)

An open standard for offering an AI agent a set of named actions it can call, such as search providers or book an appointment. More on MCP (Model Context Protocol).

Merchant of record

The business legally selling the item, responsible for payment, refunds and customer service. In most agent checkouts it stays the merchant. More on Merchant of record.

Mixed-use crawler

One crawler that does both search indexing and AI training, so refusing one used to mean refusing the other. More on Mixed-use crawler.

OpenAPI

A standard file format that describes an API so software, including agents, can read what it does. More on OpenAPI.

Pay per crawl

Charging AI crawlers a fee for each page they fetch, collected through your CDN. More on Pay per crawl.

robots.txt

A public text file at the root of a site that tells automated visitors which parts they may visit. It is a request, not a lock. More on robots.txt.

Scanner

A tool that checks a site against agent readiness practices and returns a score and failing checks. It checks files and settings, not your business decisions. More on Scanner.

Sitemap

A file that lists the pages you want automated visitors to find. Useful, though agents mostly arrive by following links. More on Sitemap.

Structured data (JSON-LD)

Facts such as price, availability and opening hours written in a machine-readable format inside the page, usually Schema.org in JSON-LD. More on Structured data (JSON-LD).

Training crawler

A crawler that takes pages to train or fine-tune an AI model. Most providers let you refuse it separately from search. More on Training crawler.

UCP (Universal Commerce Protocol)

Google's open standard, built with retailers and Shopify, that lets AI surfaces such as AI Mode and Gemini check out with a merchant. More on UCP (Universal Commerce Protocol).

User agent

The name a browser or bot gives itself with each request. Robots.txt rules and bot settings use these names, such as GPTBot or ChatGPT-User. More on User agent.

Verified bot

A bot whose operator has proven its identity to your CDN or bot vendor, through signatures, published IP ranges or a directory. More on Verified bot.

Web Bot Auth

A proposed standard in which an agent signs its requests so a site can check who operates it, rather than trusting the name it claims. More on Web Bot Auth.

WebMCP

A proposal that lets a web page tell an agent in the browser which actions it offers, as named tools, instead of the agent guessing at buttons. More on WebMCP.

x402

A way for a site to ask for payment inside the web request itself, so an agent can pay a small amount and retry without an account. More on x402.