Imagine your brand appears in nearly twice as many AI responses this month as last month. Mention rate up 22 percentage points. By every intuition carried over from two decades of search engine optimization, this is a victory. More appearances mean more visibility, more buyer exposure, more commercial impact.

The data shows the opposite can be true.

Luther and Touboul-Cohen (2026) documented nine instances across two AI platforms, ChatGPT and Google AI Overviews, where brands simultaneously gained in mention frequency while losing in positional prominence. The most pronounced example: Traditional Medicinals on ChatGPT between December 20 and January 3. Mention rate rose from 25.8% to 48.1%, an increase of 22.3 percentage points, nearly doubling the brand’s appearance frequency. Average position degraded from 3.4 to 5.9. The brand was included in substantially more responses and appeared in a less prominent position in each of them.

This is the mention-position decoupling: the finding that inclusion frequency and positional prominence in AI search are governed by separate algorithmic decisions, drawing on overlapping but distinct signals, such that optimising for one does not produce gains in the other.

For businesses that have been thinking about AI brand visibility as a single metric (are we appearing in AI responses?), the decoupling finding requires a fundamental reframe. AI brand visibility has two dimensions. Gaining on one while losing on the other is not a net positive. Both require deliberate measurement and deliberate strategy.

AI Brand Visibility: Why Appearing More Often Does Not Mean Appearing More Prominently

AI brand visibility has two independent dimensions: mention rate, how often a brand appears in AI responses, and average position, how prominently it appears when it does. The research behind this article documents brands whose mention rate rose while their position fell in the same period, so optimising for one and assuming the other follows is a mistake. The rest of the article explains what drives each and why they decouple; the broader picture is in the AI visibility guide.

What Is AI Brand Visibility and Why Does It Have Two Dimensions?

AI brand visibility is the degree to which a brand appears in, and is prominently featured within, AI-generated responses to category-relevant queries. It is measured across two dimensions that the Luther and Touboul-Cohen (2026) study establishes as operationally independent.

Mention rate is the percentage of AI-generated responses, across a defined set of category-relevant prompts and multiple independent sessions, in which a brand appears as a viable recommendation or reference. It measures breadth: in how many AI conversations about a relevant topic does this brand feature? A brand with a 40% mention rate appears in four out of ten relevant AI responses.

Average position is the mean ordinal rank at which a brand appears within the responses where it is mentioned. It measures prominence: when the brand appears, how early in the AI response does it appear? A brand consistently named in position 1 appears first, before any competitors, in every response where it features. A brand at position 4 appears after three competitors have already been named.

In traditional SEO, these two dimensions are collapsed into a single metric: ranking position. A page either ranks on the first page or it does not; if it does, its position (1 through 10) determines both whether it appears and how prominently. There is no separate “inclusion rate” that operates independently of “rank position” in traditional search.

In AI search, the two dimensions separate. The system first decides which brands are relevant to the query and belong in the response, an inclusion decision. Then it separately decides the order in which those brands appear, a prominence decision. These two decisions draw on different signals, are driven by different evaluation logic, and require different optimization strategies.

For the broader context of what AI brand visibility means as a commercial construct, see AI visibility. The discipline of generative engine optimization is specifically designed to improve AI brand visibility across both dimensions.

What Is the Mention-Position Decoupling and Why Does It Happen?

The mention-position decoupling is the finding that mention rate and average position operate independently: changes in one metric do not reliably predict or produce changes in the other.

Luther and Touboul-Cohen (2026) frame the mechanism with precision: “One is a relevance judgment. The other is a confidence judgment. They draw on overlapping but distinct signals and optimizing for one does not move the other.”

The inclusion decision, mention rate, is a relevance judgment. For each query it receives, the AI system evaluates which brands in its knowledge base are relevant options for the user. Relevance is determined broadly: does this brand exist in this category? Does it have any meaningful presence in the content ecosystem? Is there enough information about it to make it a plausible recommendation? A brand that satisfies these broad relevance criteria gets included. The breadth of inclusion, how many different query types trigger the brand’s inclusion, is what mention rate measures.

The prominence decision, average position, is a confidence judgment. Among the brands that have passed the relevance threshold, the system orders them by how confidently it can recommend each one for the specific query context. Confidence is determined by depth: how authoritative are the sources that discuss this brand? How accurately and specifically does this brand’s positioning match the exact query intent? How consistent, verified, and trustworthy are the entity signals associated with this brand? A brand that generates high confidence, through factual accuracy, expert attribution, high-authority earned media, and precise semantic fit, gets placed first.

The practical consequence of this two-decision architecture is direct. A brand can expand its relevance surface area, by producing content across more sub-topics and by increasing its topical footprint, and thereby increase its mention rate. But if that expansion comes at the cost of depth and specificity, the brand’s average confidence score across its citations may decline even as its inclusion frequency rises. This is precisely the Traditional Medicinals pattern: appearing in nearly twice as many responses but appearing less prominently in each.

For the topical authority signals that drive both breadth and depth of AI brand visibility, see topical authority SEO.

What Drives Mention Rate in AI Brand Visibility?

Mention rate is the AI brand visibility metric driven primarily by topical relevance breadth: how many different query types trigger the brand’s inclusion in AI responses.

The signals that expand mention rate are the signals that expand a brand’s relevance footprint across the content ecosystem AI systems draw from.

Topical coverage depth and breadth. A brand that has comprehensive content covering multiple sub-topics within its category creates a larger relevance surface area. For a digital visibility agency, this means content covering not just the broad category (digital marketing) but the specific sub-topics buyers ask about: AI search optimization, brand entity signals, structured data, topical authority, AI search monitoring, and every adjacent topic that generates category-relevant queries. More covered sub-topics means more query types that trigger inclusion.

Brand entity verification. Consistent cross-referenced brand identity signals (Organisation schema, Google Business Profile, NAP consistency, editorial mentions) confirm the brand’s existence and category membership for AI systems. Without strong entity signals, a brand may be topically relevant but fail the identity verification threshold for inclusion. Kargaev (2026) identifies Brand Entity Mentions as the dominant GEO signal at NIS 0.918, and entity verification is the prerequisite for any mention rate.

Content distribution. AI systems draw not just from a brand’s own website but from the full content ecosystem: editorial coverage, review platforms, industry publications, community discussions. A brand mentioned across a wide variety of sources is more likely to be triggered for inclusion across a wide variety of query types. The breadth of cross-web presence directly expands mention rate.

FAQ and conversational content. Iyappan (2026) documents that FAQ-formatted content achieves 67% AI citation rate compared to 41% for keyword-focused content, a 63% relative improvement. FAQ content maps directly onto the conversational query structure of AI interfaces, expanding the query types that trigger inclusion. Building FAQ content around the full range of buyer questions expands mention rate systematically.

For the AI content optimization research that maps citation rates by content format, see AI content optimization. The complete AI search content strategy covers the full approach to building mention rate through content investment.

What Drives Average Position in AI Brand Visibility?

Average position is the AI brand visibility metric driven by confidence: how authoritative, accurate, and semantically precise the AI system’s representation of the brand is.

The signals that improve average position are different in kind from those that expand mention rate. They are depth signals rather than breadth signals, and they require different investment types.

Factual accuracy and expert attribution. Iyappan (2026) documents that factual accuracy shows a Very Strong positive correlation with AI trust signal ratings, the highest confidence rating in the study’s correlation framework. Content with specific attributed statistics, formal citations, and verifiable expert authorship is structurally preferred by AI systems for prominent placement. A brand whose content is vague, unattributed, or anecdotally supported cannot generate the high-confidence representation that prominent AI positioning requires.

High-authority earned media. When AI systems are determining which brands to place most prominently, they draw on the authority of the sources that discuss those brands. Editorial coverage in publications that AI systems treat as highly authoritative (industry publications, respected institutional sources, quality news coverage) produces confidence signals that position a brand more prominently than an equivalent volume of lower-authority mentions. This is where digital PR connects directly to average position: a single placement in a high-authority publication produces more positional improvement than many placements in lower-authority sources.

Semantic fit with specific query intent. The Luther and Touboul-Cohen (2026) study documents that Traditional Medicinals achieves a Google AI Overviews mean average position of 1.92, the best in the dataset, despite appearing in fewer total responses than its competitors. The mechanism is semantic fit: Traditional Medicinals’ wellness positioning precisely matches the intent of herbal and wellness queries, generating high confidence that it is the right answer for those specific questions. Brands with precisely defined positioning achieve higher confidence scores for the queries that fall within their semantic territory.

Kargaev (2026) entity depth. Brand Entity Mentions at NIS 0.918 drive not just inclusion but prominent inclusion. The depth of entity verification, the richness of the cross-referenced, cross-platform brand identity that AI systems can confirm, determines how confidently AI systems name a brand specifically rather than including it ambiguously. Deep entity signals produce the confident, specific recommendations that appear in prominent positions.

For the brand entity research that explains how entity depth drives prominent AI citations, see brand entity SEO. The Google AI optimization guide addresses the content signals that Google AI Overviews uses for prominence decisions specifically.

Why Can You Not Optimise for One Metric and Assume the Other Follows?

The decoupling finding has a direct investment implication: a single-track AI brand visibility strategy is insufficient. The Traditional Medicinals case demonstrates the failure mode with empirical precision: a nearly doubled mention rate accompanied by significantly degraded positional prominence. More appearances, worse position.

Luther and Touboul-Cohen (2026) frame the strategic conclusion directly: “For marketers accustomed to optimizing a single metric, this creates a fundamental question: whether to prioritize frequency of inclusion or prominence when included, knowing that gains in one do not reliably produce gains in the other. The answer is that both require deliberate attention and progress on one cannot be assumed to produce progress on the other.”

The practical conflict between mention rate strategy and average position strategy is real. Expanding topical coverage, publishing more content across more sub-topics to increase mention rate, can dilute the depth and specificity signals that drive positional prominence. A brand that spreads its content investment across twenty sub-topics may increase its mention rate while reducing the topical depth that makes it a high-confidence citation for any specific query. The breadth-depth trade-off is genuinely in tension.

The resolution is not choosing between mention rate and average position. It is managing them as separate investment streams with separate success metrics and separate intervention logic.

The mention rate investment stream: topical coverage expansion, FAQ content architecture, brand entity breadth, content distribution and cross-web presence. Success metric: mention rate trend across multiple intervals.

The average position investment stream: factual accuracy standards, expert attribution in content, digital PR for high-authority editorial placements, precise positioning and semantic fit. Success metric: average position trend across multiple intervals on each platform.

Neither stream can be evaluated by the other stream’s metrics. A digital PR campaign that earns high-authority editorial placements should be evaluated by its average position impact, not by its mention rate impact. A topical content expansion programme should be evaluated by its mention rate impact, not by its average position impact.

For the complete AI brand visibility framework that covers both streams, see AEO vs GEO for how AEO and GEO investments map onto mention rate and average position respectively.

What Does the Decoupling Mean for B2B AI Brand Visibility?

The mention-position decoupling has specific implications for B2B businesses that are distinct from consumer brand applications.

In consumer categories, mention rate carries significant commercial value because buyers consult AI for exploration and discovery, so they may not have strong prior brand preferences, and being present in more responses increases the probability of being considered. A tea buyer who sees a brand mentioned in an AI response for the first time may be influenced by that mention even if the brand appears in position 4 or 5.

In B2B categories, the dynamic is different. B2B buyers typically arrive at AI search with at least some category knowledge and often with specific evaluation criteria. They are more likely to consult AI for validation and comparison than for discovery. In this context, appearing in a prominent position, position 1 or 2 in an AI vendor recommendation, carries more commercial weight than appearing frequently at lower positions. A B2B buyer who receives an AI recommendation that names Company A first and Company B fourth is more likely to prioritise evaluating Company A, regardless of how many total responses each brand appears in.

For most B2B service businesses, average position is the primary commercial metric of AI brand visibility, and mention rate is the secondary metric. The investment priority should reflect this: depth signals (factual accuracy, expert attribution, high-authority earned media, precise semantic fit) that drive average position should receive proportionally more investment than breadth signals that drive mention rate.

The platform specificity compounds this: Perplexity, the platform most used by professional B2B researchers, has Very High citation explicitness: it shows users which sources informed the response. Appearing in a prominent position in a Perplexity response is the AI equivalent of a prominent placement in a respected industry publication. The confidence signals that drive Perplexity prominence are the same signals that drive average position across all platforms: authority, accuracy, and semantic fit.

For the AI search platforms analysis that explains how different platforms weight inclusion versus prominence signals, see AI search platforms.

Why a narrower position can produce a better placement is examined in brand positioning in AI search.

Further reading

Frequently Asked Questions About AI Brand Visibility

What is AI brand visibility?

AI brand visibility is the degree to which a brand appears in, and is prominently featured within, AI-generated responses to category-relevant queries. It comprises two operationally independent dimensions: mention rate (how frequently the brand appears across relevant AI responses) and average position (how prominently the brand is placed within responses where it appears). Luther and Touboul-Cohen (2026) established these as distinct metrics through longitudinal analysis of 50,000+ AI responses across six real brands, documenting nine instances where the two metrics moved in opposite directions simultaneously.

Why do mention rate and average position move independently?

Because they are governed by different algorithmic decisions. Inclusion (mention rate) is a relevance judgment: does this brand belong in this response at all? Prominence (average position) is a confidence judgment: of the brands that belong in the response, which should appear first? These judgments draw on overlapping but distinct signals. Relevance signals (topical coverage breadth, entity presence, content distribution) drive inclusion. Confidence signals (factual accuracy, expert attribution, high-authority editorial coverage, semantic fit precision) drive prominence. Optimising for relevance breadth can expand mention rate without improving, and sometimes while degrading, the confidence signals that drive prominent positioning.

Which metric matters more: mention rate or average position?

The answer depends on the business type and buying context. For B2B service businesses where buyers consult AI for vendor evaluation rather than discovery, average position typically carries more commercial weight: a first-position AI recommendation drives more qualified buyer behavior than five lower-position appearances. For consumer brands seeking broad awareness across a large buyer population, mention rate matters more. For most businesses, both metrics require deliberate investment, but the proportional priority between them should reflect the specific dynamics of how buyers in the category use AI systems.

How do I improve my average position in AI responses?

Average position is driven by confidence signals: factual accuracy with attributed sources (Iyappan, 2026, Very Strong AI trust correlation), expert authorship with verifiable credentials, high-authority editorial mentions in publications AI platforms treat as authoritative sources, and precise semantic fit between brand positioning and specific query intent. Digital PR programmes targeting high-authority industry publications improve average position more directly than content volume programmes. Brand entity depth, comprehensive, cross-referenced identity signals confirming the business’s specific expertise, is the foundational average position driver (Kargaev, 2026, NIS 0.918).

Can I track mention rate and average position without dedicated software?

Yes, through manual prompt testing, but with significant limitations. Manual testing can identify approximate mention rates with consistent prompt sets run across multiple sessions. It cannot match the statistical reliability of 50-session-per-prompt methodology, and it does not scale to competitive benchmarking across multiple brands and platforms simultaneously. Dedicated tools such as Otterly.ai, Peec AI and Semrush AI Toolkit automate multi-session testing and provide trend data that distinguishes signal from noise. AIO Clicks provides AI brand visibility monitoring combining measurement infrastructure with active strategy optimisation.

How Does AI Brand Visibility Measurement Differ From Traditional Brand Monitoring?

Most organisations have some form of brand monitoring in place: social listening tools, media monitoring services, review tracking platforms. These tools measure brand presence across a wide range of sources: news coverage, social posts, review platforms, community discussions, and earned media. They are designed to answer the question: what are people saying about our brand?

AI brand visibility monitoring answers a different question: how is our brand represented in the AI-generated responses that buyers receive when they consult AI systems for recommendations, comparisons, and category guidance? The two questions overlap, since the content ecosystem that AI systems draw on includes much of the same material that traditional brand monitoring tracks, but the outputs are structurally different.

Traditional brand monitoring counts mentions: how many times was the brand mentioned, in what sentiment, in which sources. AI brand visibility monitoring measures citation decisions: in what fraction of AI-generated responses was the brand selected for inclusion, and at what position when included. The mention count in traditional monitoring reflects content volume. The mention rate in AI brand visibility monitoring reflects AI system selection logic, a deliberate algorithmic decision, not a passive occurrence.

This distinction matters for how the two types of monitoring inform strategy. A spike in traditional brand mentions, driven by a PR campaign, a news event, or a social media moment, may or may not translate into improved AI brand visibility. The AI systems may not have updated their training data or retrieval associations quickly enough to reflect the recent coverage. Conversely, a steady stream of moderate-volume editorial coverage in high-authority publications may produce AI brand visibility improvements without generating a spike in traditional brand monitoring metrics.

The most useful relationship between the two is sequential: traditional brand monitoring identifies what is being said and where; AI brand visibility monitoring identifies whether what is being said is translating into AI citation selection decisions. A brand that is generating significant editorial coverage but seeing no AI mention rate improvement may have a coverage quality issue: the publications generating the coverage are not the ones AI systems treat as authoritative. A brand with improving AI average position but flat mention volume may be seeing older, high-authority coverage generating citation confidence that newer volume-driven PR is not.

The GEO checklist covers the full monitoring framework that integrates both traditional brand presence and AI brand visibility into a coherent visibility programme.

What Do Nine Instances of Decoupling Mean Across a Ten-Week Study?

The nine documented instances of simultaneous mention rate gains and average position degradation in Luther and Touboul-Cohen (2026) represent a significant empirical contribution, but they need to be understood in the context of the study’s scope.

Nine instances across 120 data points (six brands, two platforms, two metrics, five intervals) represents a meaningful but not universal phenomenon. The decoupling occurred frequently enough to establish it as a structural feature of AI brand visibility, not an occasional anomaly, but not so frequently that every mention rate improvement was accompanied by position degradation.

What the nine instances establish is the possibility and the mechanism: mention rate and average position can and do move in opposite directions, and the cases in which they do reflect the separate algorithmic logic governing each metric. For strategy purposes, this is sufficient to require dual-metric management: the possibility that single-metric optimisation could produce the Traditional Medicinals pattern (doubled inclusion frequency, significantly degraded positional prominence) is real enough to warrant monitoring and deliberate investment separation.

It is also worth noting what the study cannot establish with these nine instances: the precise conditions under which decoupling is most likely to occur. The observational design cannot isolate the causal factors that produced the specific cases documented. Future research, particularly longitudinal studies tracking brands through deliberate content interventions, as Luther and Touboul-Cohen recommend, will establish whether specific content strategy choices reliably produce decoupling in one direction or another.

For the current state of GEO research and the distinction between what the evidence establishes firmly versus directionally, the SEO vs GEO analysis provides the most current comparative framework.

How quickly do AI brand visibility metrics respond to content investments?

The response timeline differs between mention rate and average position. Mention rate improvements from new content (expanded topical coverage, FAQ architecture, new brand entity signals) typically begin showing within four to eight weeks as AI retrieval systems index and incorporate new content. Average position improvements from confidence-signal investments (digital PR placements in high-authority publications, factual depth improvements, expert attribution) tend to develop over two to four months as the content ecosystem builds the cross-referenced authority that AI systems use for confidence scoring. Neither metric produces immediate feedback from investment; both require sustained three-to-six month evaluation windows to distinguish genuine improvement from surface volatility.

Does AI brand visibility matter for businesses not in consumer product categories?

Yes, and possibly more. The Luther and Touboul-Cohen (2026) study used a consumer product category (tea brands) because it provided a clear competitive set with established brands. But cross-industry data from the Whitebox platform, cited by the authors, indicates the five documented patterns (platform divergence, volatility, mention-position decoupling, leadership instability, and category positioning effects) appear across industry verticals. For B2B service businesses where AI-mediated vendor discovery is growing, AI brand visibility is directly connected to commercial outcomes: the brands that appear prominently in AI-generated vendor recommendations are entering buyer consideration sets before any direct commercial contact occurs.