Terms and metrics used across LLM Authority Index.

The term is broader than a simple visibility metric. It describes the research hypothesis that persistent changes in AI recommendation behavior may contain information about future consumer consideration, branded search, direct traffic, customer acquisition, or other commercial outcomes. That relationship is being tested prospectively and has not been established as a predictor of revenue, earnings, valuation, or stock returns.

Under the current V0 framework, a Positive AI Divergence Candidate requires a recommendation-coverage increase of at least 5 percentage points, an exploratory interval above zero, and improvement across at least four AI platform families. A Negative AI Divergence Candidate requires the reverse pattern. The label is a research classification, not a stock rating or investment recommendation.

A change in AI visibility is not automatically an investor signal. In the LLM Authority Index framework, investor relevance must be tested prospectively. A useful signal should be measured before the downstream outcome, economically connected to the company or tracked entity, reasonably persistent, and incrementally informative beyond conventional data investors already have.

The reports treat answer position as commercially meaningful because AI-generated recommendations are not neutral lists; users tend to notice and trust the first few names most. The system determines ranking by looking for an explicit ordered list or "best / top / first / second" structure, and if none exists, it defaults to first tracked-company mention order. This is included because being ranked first, second, or third has different likely outcomes than being buried lower in the response, even when overall visibility looks similar.

In the longitudinal AI Investor Signals framework, recommendation coverage is generally calculated across matched prompt-platform observations. It measures absolute recommendation frequency, not simple presence and not competitive recommendation share. A company can appear in an answer without being recommended, and it can increase recommendation coverage while losing recommendation share if competitors improve faster.

In the current LLM Authority Index framework, the primary implementation measures the change in recommendation coverage across matched prompt-platform observations. AI Recommendation Momentum is an AI-side measurement. It should not be interpreted as revenue momentum, earnings momentum, stock-price momentum, or financial performance without separate validation.

Recommendation share measures competitive selection, while recommendation coverage measures how often the company is recommended across eligible observations. The two can move differently. A company can improve absolute recommendation coverage while losing recommendation share if competitors improve faster.

AI Recommendation Share Gap = AI Recommendation Share - Comparable Market Share. A positive gap means the company captures a larger share of AI recommendations than its current real-world market position would suggest. A negative gap means it captures a smaller share. The gap is descriptive until longitudinal research establishes whether it contains information about future commercial share or growth.

This is the boardroom-friendly expression of the model: how much AI-influenced demand value a brand appears to control. It is not meant as exact attribution, but as a disciplined estimate of the revenue pool associated with AI recommendation share. It is included because it gives the report a commercially legible output rather than stopping at abstract visibility numbers.

This is one of the foundational metrics in the report because it answers the most basic discovery question: when people ask commercially relevant questions in AI tools, how often does the brand show up at all? It is usually calculated across a defined prompt set, cluster, platform, or total market view. Its value is that it gives a directional market-share style view of AI visibility, but it is included with an important caveat: appearance alone is not enough, because a brand can be visible without being preferred, recommended, or ranked highly. That is why SOV is always interpreted alongside ranking, recommendation rate, and citation strength.

The term does not mean that the market is wrong or that a stock is mispriced. A divergence only identifies a potential difference between an AI-derived commercial signal and expectations already visible in financial data. The AI signal must first demonstrate prospective commercial value before market-divergence conclusions become meaningful.

A brand's overall sentiment may look healthy while specific attributes are weak or risky. For example, a platform may be described positively for ease of use but negatively for pricing transparency. This is included because strategy usually requires knowing what exactly is helping or hurting recommendation quality, not just whether the brand is liked in general.

The scoring model assigns 10 points to rank 1, 9 to rank 2, and so on down to 1 point for rank 10, with rank 11+ scoring zero. This allows the reports to compress many prompt outcomes into one comparable ranking measure. It is included because raw rank snapshots are too fragmented at scale; Average Rank gives a clearer view of how strongly a brand performs across an entire cluster or platform rather than in a single answer.

This isolates answer quality from answer frequency. A company might have low overall visibility but rank very strongly whenever it does appear, or it might appear often but only in low positions. This metric is included to separate those cases and prevent analysts from confusing absence problems with ranking-quality problems.

This version of AR divides total ranking points by all prompts in the cluster, so absence hurts the score. That makes it a better measure of overall cluster strength than a "when mentioned" score. It is included because it answers the tougher question: not just how well the company ranks when it shows up, but how strong it is across the full opportunity set.

It is a proposed framework for measuring how strongly a brand is positioned within the network of sources, claims, and evidence that influence AI-generated answers and recommendations.

A company often performs better when it is already named in the question, but the more valuable discovery signal is whether it appears organically in non-branded prompts. This is included because branded presence measures existing awareness, while organic appearance measures real recommendation power and category relevance.

The report architecture recognizes that prompts at different stages have different commercial implications. Early educational prompts shape category entry, while pricing and shortlist prompts influence conversion-ready decisions. Buyer Stage is included because the same visibility level can mean very different things depending on whether the user is just learning or about to choose.

This is one of the most important explanatory concepts in the report. AI models do not form recommendations out of nowhere; they are influenced by the domains, content types, and third-party references that repeatedly support answers. Citation Architecture captures those support structures, including official sites, editorial sources, reviews, forums, and other public references. It is included because many ranking and recommendation outcomes are driven less by brand messaging and more by the surrounding evidence environment.

The sample reports repeatedly show that some competitors win not just because of brand strength, but because they have built an evidence environment AI systems repeatedly rely on. That produces self-reinforcing recommendation patterns. It is included because it explains why some brands are hard to displace even when the target has a competitive product.

It is a proposed framework for measuring how influential a source is within the evidence networks that shape AI-generated answers, recommendations, and conclusions.

The methodology specifically calls for classifying citations by source type because not all evidence plays the same role. Official pages may explain products, editorial content may define "best" lists, reviews may shape trust, and forums may shape real-user credibility. It is included because understanding the mix helps explain why certain brands are framed as authorities while others are treated as alternatives or cautions.

Tracking cited domains shows which websites are repeatedly used to support brand recommendations, explanations, or comparisons. This matters because the source of the answer often influences the framing of the answer. It is included to reveal whether the brand is supported by strong first-party and third-party evidence, or whether competitors own the domains AI systems keep leaning on.

This metric goes beyond overall domain counts and asks which sources are effectively carrying the brand's authority. That can include the brand's own website, partner pages, editorial reviews, public discussions, or comparison sites. It is included because a company may appear often, but for fragile reasons tied to a narrow evidence base, while another brand may be supported by a broader, more durable citation network.

The report is explicitly competitive intelligence, so it does not stop at describing the target in isolation. It compares where competitors outrank, out-cite, or out-convert the target in commercially important conversations. This is included because strategy depends on understanding not just internal weakness, but who is taking the recommendation share the target is losing.

This turns momentum into a comparative signal. Instead of asking only whether the target improved, it asks whether the target improved faster or slower than the brands it competes against. It is included because a company can post positive gains and still lose relative position if competitors are accelerating faster.

Higher CPC often signals that advertisers value traffic in that area, which makes it a useful proxy when estimating the economic importance of prompt clusters. It is included because the report's economic layer needs a directional commercial weighting system grounded in something more concrete than opinion.

This is the value layer that converts discovery metrics into business relevance. Rather than treating appearance in AI answers as a vanity signal, the economics layer connects visibility to directional monetary value using search demand, commercial intent, and monetization proxies. It is included because decision-makers care not just who appears in AI answers, but what those appearances are likely worth.

This is one of the more sophisticated parts of the methodology because it captures market role, not just tone. A brand may receive positive sentiment but still be framed as niche, secondary, or conditional. It is included because the reports are meant to explain not just whether AI mentions a brand, but what kind of market position AI seems to believe that brand occupies.

Examples include pricing, alternatives, trust, educational, or recruitment software prompts. Clustering matters because discovery performance varies dramatically by intent type; a brand may be strong in alternatives but invisible in pricing, or trusted in educational content but absent from shortlist prompts. Clusters are included because they make the report strategically useful: instead of one blended number, the report shows where the commercial problem actually lives.

Matched cells are used to reduce denominator drift in longitudinal AI measurement. Instead of comparing different prompt or platform populations from month to month, the methodology compares the same normalized prompt on the same platform family when possible. This makes measured momentum less sensitive to changes in the observation mix.

This is a conversion metric rather than a visibility metric. It asks: once the brand enters the answer, how often does it become the leading recommendation? That matters because some brands are commonly present but rarely endorsed as the best option. It is included because it reveals the gap between being known and being preferred.

This is similar to Mention-to-Top-1 but slightly broader and often more stable across clusters. It is useful for diagnosing whether a brand is merely peripheral or genuinely considered competitive once it is surfaced. It is included because a company can have acceptable presence but poor conversion into commercially meaningful positions.

The system is designed for recurring reporting, so one static snapshot is not enough. Momentum shows whether share of voice, ranking, citations, and other indicators are improving, weakening, or holding flat. It is included because early movement often matters more than current size; a smaller brand gaining quickly may be strategically more important than a larger brand standing still.

The label means the AI-side movement is unusual enough to merit prospective follow-up. It does not mean the company's revenue, earnings, valuation, or stock price will decline.

Platform breadth is used as a separate signal-quality dimension because aggregate recommendation change can hide disagreement among AI systems. Broader directional participation may make an observed AI-side movement more robust, but it does not by itself establish commercial or financial predictive value.

A company can show strong aggregate improvement even when most platforms are flat or declining if one platform moves sharply enough. Platform concentration risk therefore helps distinguish broad multi-platform movement from platform-specific movement. It is a measurement-quality concept, not a claim about the company's financial risk.

The methodology explicitly separates datasets by platform because the same brand can perform differently across LLM environments. Each platform has its own retrieval patterns, answer structures, and citation tendencies, so a brand may win on one system and disappear on another. It is included because "AI visibility" is not one market; it is a set of overlapping but distinct discovery environments.

Because AI systems evolve quickly, visibility and recommendation behavior can shift by platform and by month. Tracking volatility helps analysts avoid overinterpreting one-off wins or losses. It is included because a durable opportunity is more valuable than a temporary fluctuation caused by platform instability.

The label means the AI-side movement is unusual enough to merit prospective follow-up. It does not mean the company's revenue, earnings, valuation, or stock price will increase.

Presence is a visibility measure rather than an endorsement measure. A company may be mentioned as background context, compared with competitors, described negatively, or listed without being recommended. Presence should therefore be interpreted separately from recommendation coverage, recommendation share, rank, sentiment, and citation behavior.

The reports are built around the idea that not all prompts matter equally; what matters is coverage across high-intent questions that align with buying, comparing, trusting, or shortlisting. Prompt Coverage therefore shows where the brand is active and where it is absent across actual demand. It is included because brands often discover that their AI visibility is patchy: strong in some buyer questions and nonexistent in others.

Within a larger cluster like "trust" or "comparison," prompts can still behave differently depending on wording, user need, or category angle. Classifying subtypes helps isolate which exact conversations the brand wins or loses. It is included because broad clusters can hide important nuances that matter for strategy and content intervention.

Intent classification helps keep the report commercially grounded. It explains why some prompts are informational, some are evaluative, and some are directly transactional in nature. It is included because the report's usefulness depends on distinguishing curiosity from buying behavior.

In the economics model, not all prompts are equally valuable; some represent much larger pools of demand than others. Query Volume is included because a brand winning a low-volume prompt cluster is less meaningful than winning a high-volume cluster with strong buyer intent. It helps weight visibility by actual market demand.

If recommendation coverage moves in the same direction across most major AI systems, the movement is more portable. If aggregate movement is driven by only one or two platforms while others disagree, portability is lower. Recommendation portability measures cross-platform consistency, not financial strength or predictive validity.

Search Volume is often used as an input into cluster weighting and economics modeling because it helps quantify the size of the opportunity behind specific themes. It is included because the report aims to focus on real demand, not just arbitrary prompt samples.

The methodology treats sentiment as contextual rather than simplistic. It considers the user question, the wording used by the AI, and the role the company plays in the answer. This is included because a brand can be visible but framed negatively, qualified cautiously, or treated as a weaker option. Visibility without favorable sentiment may not translate into trust or conversion.

Rather than looking only at raw counts, these summary measures provide a quicker way to compare how a brand is being described across clusters or platforms. They are included because analysts need a compact read on whether brand framing is improving, deteriorating, or holding steady over time.

These are some of the most actionable records in the report because they represent live discovery losses. They show the exact moments where the market is having a conversation and the target brand is excluded from it. They are included because absence is often more strategically revealing than weak presence.

This is the clearest measure of recommendation leadership. It shows how often the brand is the default, primary, or "best" answer rather than just one option among many. The value of Top-1 Rate is that it captures who is truly winning the buying moment. It is included because many brands are visible but rarely win the top slot, which creates a structural disadvantage that basic visibility metrics can hide.

This metric extends the ranking lens further down the answer to show whether a company is broadly present in ranked results even when it is not near the top. It is most useful for large comparison-style prompts where more brands appear. It is included because it helps distinguish total invisibility from weak-but-real inclusion and supports the ranking score logic used in Stage 0 extraction.

This is a practical conversion metric because many users focus on the first few options in a list. A brand that consistently lands in the top three is still meaningfully in the consideration set, even if it is not always first. It is included because it provides a more forgiving but still commercially useful measure of competitiveness than Top-1 alone, especially in crowded categories where the brand may not dominate but still earns serious evaluation.

These are the openings the report is meant to surface: clusters, platforms, or citation environments where the target is not yet strong but where recoverability or upside looks favorable. It is included because the report is not just diagnostic; it is supposed to show where gains are most realistic and commercially meaningful.

VPQ usually comes from affiliate economics, monetization benchmarks, CPC proxies, or similar commercial inputs. It translates demand into value instead of treating all queries equally. It is included because some prompt categories are much more monetizable than others, and the report's economics layer depends on reflecting that uneven value distribution.

The automation specification includes this to prevent all clusters from being treated as equally important. By combining demand and commercial signals, the report can prioritize discovery areas that likely matter more to revenue or lead generation. It is included because strategic recommendations should be guided by opportunity size, not only by visibility weakness.