This glossary defines the terms that actually matter in Generative Engine Optimization (GEO) — the practice of getting a business mentioned, recommended and cited inside AI answers from ChatGPT, Claude and Perplexity. Every entry below is a term used elsewhere on this site and grounded in the engines' own documentation, not marketing jargon invented to sound technical. If you're new to GEO, start with entity and Buyer Question — those two decide most of what an engine can even consider recommending you for.
Generative Engine Optimization (GEO) is the practice of making a business visible and recommendable inside AI-generated answers — in ChatGPT, Claude and Perplexity — the way SEO made a business visible in Google's ranked links. The full walkthrough of what that involves, how it differs from SEO and what actually moves the needle lives on the what is GEO pillar page; everything else in this glossary is a term used somewhere in that chain.
Bottom line: GEO doesn't replace SEO, it sits on top of it — the terms below describe the layer above ranking, not a replacement for it.
An entity, in GEO terms, is the single identity an AI engine resolves your company name, website and social profiles into before it will recommend you. Naming a business in an answer is a claim about a specific real-world entity, not just a link to a page — so the engine first checks whether your domain, your company name and your LinkedIn page collapse into one confirmed identity. Getting this right is what entity clarity means in practice: one consistent name, a real About page, and profiles linked together.
Do this: before touching page content, confirm an engine can resolve who you are at all — content quality only matters after identity is settled.
Independent corroboration is a claim about your business appearing on a site you don't control — a review, a comparison article, a forum thread — which is what AI engines weigh as evidence when deciding whether to trust a claim your own site makes about itself. A page that says "the fastest CRM setup" is a claim; a third-party review repeating that claim independently is corroboration.
Proof: Google's guidance on generative AI features names firsthand experience and independent corroboration as signals it looks for, and explicitly warns against manufacturing mentions to fake it — Google — optimizing for generative AI features.
Rule of thumb: you can't buy corroboration convincingly — earn it through real reviews, comparisons and mentions instead.
An AI crawler is the bot each engine sends to fetch and read pages before it can cite them — GPTBot and OAI-SearchBot for ChatGPT, ClaudeBot for Claude, Google-Extended for Gemini and AI Overviews, and PerplexityBot for Perplexity. Block one, even by accident through a CDN's default bot-blocking rule, and that engine goes dark on your business regardless of how good your content is.
Proof: OpenAI's own documentation states that appearing in ChatGPT search requires OAI-SearchBot not being blocked at the robots or host level — OpenAI — ChatGPT Search.
Do this: check your robots.txt for each AI crawler by name — a wildcard rule meant for something else can block one without anyone noticing.
Answer-first writing is a content structure that states the direct answer to a question in a section's first sentence, before any context, backstory or caveats. AI engines quote passages that answer a question cleanly — a page that circles the topic for three paragraphs before getting to the point rarely gets extracted, even when the eventual answer is correct. The full pattern, with before/after examples, is on the answer-first writing page.
Shortcut: read your own first sentence under each heading — if it isn't the answer, move the answer up and push the context down.
A Buyer Question is the actual question a prospective customer types or speaks to an AI engine while deciding what to buy — "best CRM for a 10-person team," not a keyword like "CRM software." It's the unit of analysis Reflexa audits are built from: a set of real buyer questions per category, run against ChatGPT, Claude and Perplexity, with the answers scored for what actually comes back.
Do this: write down the exact questions your last five customers would have asked an AI before they bought — that list is worth more than a keyword tool.
Mentions, Recommendations and Citations are the three things Reflexa counts when it measures AI visibility. A Mention names your business anywhere in an answer; a Recommendation names you specifically as a good answer to a Buyer Question; a Citation links to or names your site as the source behind a claim. They're reported as observed counts per engine — never rolled up into a single made-up score.
Bottom line: a Mention without a Recommendation means the engine knows you exist but isn't ready to suggest you — that's usually an entity or corroboration gap, not a content one.
The free check runs in about three minutes, with the raw evidence behind every finding.