Home › Glossary

Reference

AI visibility, in plain English

The AI-visibility field has grown a thicket of acronyms fast. Here are the 17 terms worth knowing — what they mean, why they matter, and no more jargon than necessary.

AEO (Answer Engine Optimization)

The practice of making your business appear in the answers that AI assistants and answer boxes give, rather than in a list of links. Where SEO optimizes for ranking on a results page, AEO optimizes for being named or cited inside a synthesized answer. In practice it spans on-site work (clear, structured content) and off-site work (the third-party sources AI engines trust).

GEO (Generative Engine Optimization)

A near-synonym of AEO, focused specifically on generative AI engines like ChatGPT, Claude, Gemini, and Perplexity. The term comes from academic work on how content properties affect inclusion in LLM-generated answers. If you hear AEO and GEO used interchangeably, that's because in day-to-day practice they largely are.

LLM citation

A source link that an AI assistant attaches to a claim in its answer — the small footnotes or reference cards you see under a ChatGPT or Perplexity response. Citations reveal which surfaces the engine actually trusted to build its answer, which makes them the single most useful diagnostic signal in this field. Notably, they're often not the recommended business's own website.

llms.txt

A proposed standard: a plain-text file at the root of your website that gives AI crawlers a curated, machine-readable summary of what your site contains, similar in spirit to robots.txt. Adoption by crawlers is still early — a May 2026 million-site analysis found 97% of llms.txt files received zero AI-crawler requests — so it's best treated as cheap, forward-looking hygiene rather than a growth lever.

Structured data / schema markup

Machine-readable labels (usually JSON-LD following the schema.org vocabulary) embedded in a web page that state facts explicitly: this is a law firm, at this address, offering these services, with these reviews. Humans never see it; retrieval systems and search engines consume it directly. It removes guesswork when an AI engine verifies who you are and what you do.

Answer engine

Any system whose output is a direct answer rather than a list of links — ChatGPT, Claude, Gemini, Perplexity, and Google's AI Overviews all qualify. The defining commercial property is compression: an answer engine names perhaps three to six businesses where a search page listed ten or more, so inclusion is closer to binary than ranked.

AI Overviews

Google's AI-generated summary block that appears above traditional search results for many queries. It answers the question directly on the results page, which sharply reduces clicks to the links below it — Ahrefs measured a ~58% drop in click-through for the #1 organic result when an Overview is present. It's the clearest example of AI answers overwriting search behavior inside search itself.

Retrieval

The step where an AI assistant goes and fetches source material — search results, directory pages, reviews, forum threads — before writing its answer. If your business never gets retrieved for a question, you cannot appear in the answer, no matter how good you are. Most AI-visibility work is ultimately about becoming retrievable for the questions your prospects ask.

Grounding

The process of tying an AI answer to verifiable sources, so the model asserts only what its retrieved material supports. Grounding is why consistency matters: if directories, reviews, and your website disagree about your address or specialty, you become risky to cite and tend to get dropped. Well-grounded engines like Perplexity will cite as they write; others ground more loosely.

Hallucinated recommendation

An AI answer that names a business, detail, or attribute that isn't supported by any real source — a firm that doesn't exist, a wrong phone number, a specialty you don't have. These arise when a model fills gaps in weak retrieval with plausible-sounding text. They're a real reputational hazard, and one reason to periodically check what AI engines actually say about you.

Share of voice in AI

Of all the times AI engines answer a set of relevant prospect questions, the percentage of answers in which your business is named — usually compared against competitors on the same questions. It's the AI-era equivalent of rank tracking, adapted to a world where answers vary by engine, phrasing, and day. Measured well, it requires asking the same question set repeatedly across engines.

Prompt fan-out

The technique where an AI assistant silently expands one user question into several internal search queries — variations, sub-questions, and related angles — and retrieves results for each before composing a single answer. It explains why long-tail, specific content gets cited even when nobody types those exact words. Your content competes against the fan-out, not just the visible question.

Mention vs. citation

A mention is when the AI names your business in its answer text; a citation is when it links a source. You can be mentioned without being cited (the model knows you from training data or uncited retrieval) and cited without being mentioned (your page supported a general claim). Mentions win you customers; citations tell you which surfaces earned the mention.

Fix pack

Citewise's term for the remediation bundle that follows a diagnostic: the concrete, prioritized set of changes — llms.txt, schema markup, content and consistency fixes — addressing the specific gaps the measurement found. The general principle travels beyond us: diagnose first, then fix the gaps you actually have, rather than applying generic checklists.

Visibility delta

The change in your AI visibility between two measurements — before and after fixes, or quarter over quarter. Because AI answers are probabilistic and engines refresh constantly, a single snapshot can mislead; the delta across consistent, repeated measurement is the honest signal of progress. It's the number a monitoring cadence exists to produce.

Agentic search

When an AI system doesn't just answer but acts: comparing providers, filling in a booking form, or completing a purchase on the user's behalf. It raises the stakes of AI visibility, because the assistant may go from recommending a shortlist to selecting a single winner and transacting with it. Early today, but the direction of travel for high-intent queries.

Zero-click

A search or AI interaction where the user gets what they need without visiting any website. AI answers make zero-click the default for informational queries, which erodes the traffic-based metrics businesses have relied on for decades. The strategic response is to measure presence in the answer itself, since the answer is increasingly where the decision happens.

AI crawler

A bot operated by an AI company that fetches web pages to feed model training or live retrieval — examples include GPTBot (OpenAI), ClaudeBot (Anthropic), and PerplexityBot. Whether you allow them in robots.txt determines whether those systems can read your site at all. Blocking them protects content; it also removes you from the sources those engines can cite.

Want to see these concepts applied to a real business? Read how AI assistants actually build shortlists or browse a sample Citewise report.