Glossary
The vocabulary,
defined.
A young category generates a lot of terminology, some of it meaningful and some of it invented to differentiate identical services. These are the terms that carry real distinctions.
21 terms
01Terms
- Answer Engine Optimization (AEO)See also: GEO, LLMO
- The practice of structuring content, entity signals and third-party corroboration so that AI systems cite and recommend a brand. AEO competes for inclusion inside a single synthesized answer rather than for position on a results page.
- Generative Engine Optimization (GEO)See also: AEO
- Functionally synonymous with AEO. The term originated in academic work by Aggarwal and colleagues at Princeton, published at KDD 2024, which introduced the GEO-bench dataset and found that adding quotations, statistics and citations can raise generative-engine visibility by up to 40%.
- Large Language Model Optimization (LLMO)
- A third label for the same discipline, framing the language model rather than the interface as the optimization target. Tends to be used by practitioners emphasizing training-data presence over live retrieval.
- Citation
- An outcome in which your domain appears as a linked source within an AI-generated answer. The most valuable of the three visibility outcomes because it produces referral traffic as well as credibility.
- Mention
- An outcome in which your brand is named in an AI answer but no link is provided. Valuable for familiarity and often a precursor to citation, but generates no traffic — which is why it should be tracked separately.
- Recommendation
- An outcome in which a brand is named as a preferred option rather than merely referenced. The hardest outcome to earn and the one most directly tied to revenue. A brand can be cited frequently as a source while never being recommended as a choice.
- Prompt-space mapping
- The AEO equivalent of keyword research. Rather than short queries, it inventories the full conversational questions buyers ask AI systems, spanning category-definition, comparison, objection and purchase-intent prompts.
- Semantic chunk
- A passage of content that remains coherent and accurate when extracted from its surrounding page. Because retrieval systems return passages rather than documents, chunk-level self-containment determines extractability.
- Answer-first structure
- A content pattern in which a heading is phrased as the question and a direct answer of roughly forty to fifty words appears immediately beneath it, before elaboration. Optimizes for passage extraction.
- Entity resolution
- The process by which an AI system determines that a set of references — a website, a LinkedIn page, a directory listing, a news mention — all describe the same real-world organization or person. Low entity confidence suppresses recommendation.
- sameAs
- A Schema.org property declaring that an entity is identical to an entity described at another URL. Frequently the highest-leverage structured-data correction available, because it consolidates scattered weak profiles into one well-corroborated entity.
- Consensus signal
- A third-party source that corroborates a brand's claim to authority — press coverage, community discussion, reviews, partner references, conference programs. Because models reproduce aggregate impressions, consensus signals decide recommendation outcomes.
- Citation decay
- The loss of a previously held AI citation. Between 40% and 60% of AI citations change monthly, so decay is expected rather than exceptional and is used operationally as a content-refresh trigger.
- Live retrieval
- An answer-generation mode in which the engine queries the web at the moment of the question. Newly published content can appear within days, making crawler access and render path the binding constraints.
- Training absorption
- The incorporation of content into a model's weights during training. It influences answers without live retrieval but cannot be updated on demand, changing only at the next model release.
- Render path
- How content reaches a client. Server-rendered content exists in the initial HTML response; client-side-rendered content appears only after JavaScript executes. Many AI crawlers do not execute JavaScript, making render path a common cause of invisibility.
- llms.txt
- An emerging convention: a plain-text file at a site's root describing the organization and its key pages in machine-readable prose, intended to help AI systems understand a site without inference.
- GPTBot
- OpenAI's training-data crawler. Distinct from OAI-SearchBot, which powers ChatGPT search retrieval, and ChatGPT-User, which fetches pages when a user prompt triggers browsing. Blocking GPTBot alone removes a site from future training data while leaving live citation intact.
- Zero-click search
- A search session ending without any click to an external site. Bain found roughly 80% of searchers rely on zero-click results for at least 40% of searches, with associated organic traffic falling 15% to 25%.
- Reputation engineering
- The practice of monitoring and correcting what AI systems say about a brand, combining answer-engine sentiment tracking with source-level correction, review systems and entity management. Distinct from traditional ORM because the outputs are private and leave no auditable public trace.
Next step
Tell us what you are building.
If the question is about discovery, reputation, a new site, or a market move, send the context. AIGNCI will tell you whether an Audit, a build, or a more focused engagement is the right starting point.