Glossary
Working definitions for every term in the RankCaster vocabulary. Cite these in client decks, slide them into RFPs, link directly to a definition with /glossary#apr.
The share of AI answers that mention a brand for a given prompt, averaged across runs.
APR measures how often a brand appears in the body of an LLM answer for a tracked prompt. RankCaster runs each prompt 10× per cycle and reports APR with a confidence interval (e.g. APR 45% ±8pp) so the number survives the non-determinism of LLM outputs.
Marketing discipline focused on getting a brand cited by LLMs in the answers users actually see.
AVM is the umbrella discipline that contains GEO and AEO. It treats the LLM answer as the surface that matters, not just the link in a SERP. Practiced as a service line by SEO and GEO agencies.
Optimization for generative LLMs (ChatGPT, Claude, Gemini) where the brand has to appear inside the answer.
GEO is the technical and content discipline that prepares a site for citation by generative LLMs. Different ranking signals from SEO — structured data, llms.txt, MCP servers, canonical-claim assertions, entity graph completeness — and a different success metric (APR vs SERP position).
Optimization for answer engines (Perplexity, Gemini answers) where appearing in the cited sources is the goal.
AEO targets engines that explicitly cite their sources. Structured data quality, schema.org type coverage, AI-native types (Dataset, DefinedTerm, ClaimReview), and high-authority outbound citations are the levers.
Foundation layer — robots.txt, llms.txt, and AI crawler accessibility — that lets AI models read the site at all.
AIO is the bottom of the stack. If GPTBot / ClaudeBot / Google-Extended can't crawl the site, no GEO or AEO work matters. Free AI-readiness audit on /audit scores AIO first.
Open protocol that lets AI agents query a brand's data directly through a server the brand controls.
MCP lets you publish a server at /.well-known/mcp that AI agents can call to fetch authoritative facts about the brand. RankCaster ships MCP-as-a-Service — a hosted MCP server in five minutes per brand, no engineering ticket.
Bundle of machine-readable brand data — schema.org, entity graph, llms.txt, MCP discovery — shipped as real files.
LLM-pack is the build-side output of RankCaster. Schema.org structured data (Organization, Product, FAQPage, Article, DefinedTerm, Dataset), an entity graph linking the brand to canonical sources, llms.txt + llms-full.txt per spec, MCP discovery at /.well-known/mcp, and canonical-claim assertions. Files you drop into the site.
Revenue and pipeline attributed back to AI-referred sessions, measured per LLM and per prompt.
ROAI ties Answer Presence Rate to dollars. Sessions referred from ChatGPT, Gemini, Perplexity, or Claude land in GA4 with UTM tagging, conversions are attributed back to the specific LLM and prompt, and the result is a revenue number per prompt cluster.
