AI Reputation Management
The discipline of measuring, defending, and shaping what AI engines say about a company, founder, or product. Distinct from PR, SEO, and online reputation management. Combines audit methodology, earned media, Wikipedia work, structured owned content, and crisis protocols built for AI-mediated incidents.
AI-held Reputation
The version of an entity's reputation that exists inside AI engines — formed by the citation graph the model retrieves from, weighted by source authority, and compressed into a synthesized answer. Often differs from the entity's real-world reputation in instructive ways.
Citation Share
How often a brand appears when an AI engine answers a category-defining prompt across the five major engines. The AI-era equivalent of market share. Forms faster, decays faster, and rewards a different stack of inputs than traditional market share.
Authority Stack
The hierarchy of sources AI engines tend to weight when forming an answer. From highest to lowest typical weight: Wikipedia; major news outlets; mid-tier and specialist outlets; original research and institutional sources; structured owned content; social and forum content; SEO content farms and low-authority blogs.
The Five Dimensions of AI Reputation
The five axes along which AI reputation is measured: Accuracy, Sentiment, Completeness, Consistency, Control. Composite score gives the headline. Dimension breakdown gives the repair plan. Used as the scoring framework in 5W's Reputation Index.
Accuracy
Dimension of AI reputation measuring whether the model states correct facts about the entity. Tests founding year, leadership, products, financial figures, and other verifiable detail. Fails most often due to thin authority stack, conflicting sources, or stale dominance of out-of-date references.
Sentiment
Dimension of AI reputation measuring the narrative framing the model summarizes — favorable, neutral, or hostile. Skews when one critical source becomes the model's dominant reference. Repaired by displacing weight, not deleting sources.
Completeness
Dimension of AI reputation measuring whether the model's compressed summary cites the entity's key strengths or only the weak narrative. Gaps are usually the result of a thin authority stack — the strengths exist but weren't written into the sources the model pulls from at sufficient density.
Consistency
Dimension of AI reputation measuring whether the answer holds across the five major engines or fractures by platform. Inconsistency is its own reputation risk — buyers cross-check, and a fractured picture often reads as untrustworthy.
Control
Dimension of AI reputation measuring the brand's ability to influence future answers through earned media, Wikipedia, structured owned content, and the citation graph the model retrieves from. Never absolute. Measured as the share of the model's cited source pool that the brand can shape versus inherit.
Retrieval Layer vs Search Layer
The structural shift from a web mediated by search engines (returning ranked links for the user to click) to a web mediated by retrieval engines (returning synthesized answers the user reads directly). The retrieval layer is the foundation of AI reputation — the surface where brands are evaluated before any human click.
Prompt-to-Entity Association
The strength of the model's learned association between a category prompt (e.g., "best CRM platforms") and a specific brand. A brand can be well-known and still have weak prompt-to-entity association — the model knows the brand exists but doesn't surface it for the category. Fixed by repetition: naming the brand alongside the category in trusted sources, at volume.
Anti-hallucination Floor
The base layer of structured, consistent, authoritative content — owned and earned — that prevents the model from inventing detail about an entity. When the floor is missing, the model fills gaps with statistically plausible but incorrect content. Building the floor is a foundational AI reputation discipline.
Citation Graph
The full network of sources an AI engine has been trained on and retrieves from, plus the authority weights it assigns to each. The citation graph is the substrate of AI reputation. Editing the citation graph is the mechanism by which brands change AI outputs.
Live Retrieval
The process by which AI engines fetch current web content at the moment of a user prompt, in addition to or instead of relying on frozen training data. Perplexity is closest to pure live retrieval. ChatGPT with browsing, Google AI Overviews, and Gemini use it in varying degrees. Claude uses it when browsing is enabled.
Training Corpus
The frozen body of text — public web, books, licensed content — an AI engine was trained on. Sets the baseline picture the model has of any entity before any user prompt. Updated only at major version releases. The "floor" of AI reputation; live retrieval can move the model above it.
Compression Failure
A failure mode in which the model picks the wrong attributes to summarize when compressing the candidate pool into an answer. Distinct from hallucination — the underlying facts may be correct, but the model has surfaced the weakest or least relevant of them. Often caused by uneven authority weighting across the source mix.
Hallucination (AI Reputation Context)
The model's invention of plausible-sounding detail when the retrieved evidence is thin, conflicting, or missing. In AI reputation, hallucinations about brands typically trace to one of three causes: thin authority stack, conflicting sources, or stale dominance. Reduced by lock-down of high-stakes facts in tier-1 sources, complete Wikipedia presence, and structured owned content.
The Five Engines
The five major AI engines through which AI reputation is mediated and measured: ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Perplexity, and Google AI Overviews. Each weights sources differently, updates on different cadences, and pulls from different mixes. A complete AI reputation audit tests all five.
Generative Engine Optimization (GEO)
The discipline of optimizing communications, content, and authority infrastructure for citation by generative AI engines. The successor to SEO in the retrieval-layer era. Targets the citation graph the engines retrieve from rather than the link rankings of traditional search. A core function of AI Reputation Management.
Reputation Index
5W's recurring research franchise auditing AI-held reputation for brands and public figures across the five engines. Scored along the five dimensions of AI reputation using a reproducible methodology. The methodology layer that converts the framework of AI reputation into measurable, repeatable research.
Composite Score
The single headline number summarizing an entity's AI reputation across the five dimensions. Provides directional view. The dimension breakdown — not the composite — is what drives the repair plan.
Source Weight
The relative authority an AI engine assigns to a given source when constructing an answer. Weighted by domain authority, recency, citation density, editorial signal, and schema. A Reuters citation typically carries materially higher source weight than a low-authority blog citation for the same fact.
Original reporting from the top tier of news outlets — globally: The New York Times, The Wall Street Journal, Reuters, Bloomberg, Financial Times, The Washington Post, Associated Press. Category-specific: the leading trade and specialty publication for the brand's vertical. The category of source the AI engines tend to weight most heavily after Wikipedia.
No communications firm can guarantee specific outputs inside third-party AI systems. The discipline is shaping the inputs the engines retrieve from — not directing the engines themselves.
