What 126 Million AI Prompts Reveal About the Gap Between Visibility and Citation Authority

Semrush's 2026 AI Visibility Index analyzed 126 million U.S. AI search prompts across ChatGPT, Gemini, Google AI Mode, and Google AI Overviews — scaling from 2,500 prompts in September 2025 to the largest public dataset on brand presence in AI answers. The data confirms what the Machine Relations Index measures at source level: AI engines disagree fundamentally about which brands to cite, name, and surface. But the Semrush dataset measures brand visibility, not citation authority — and the structural gap between those two concepts explains why 45% of marketing leaders still cannot accurately track their brand's performance in AI answers.

Last updated: June 28, 2026

What the 126 Million Prompt Dataset Actually Measures #

Semrush's expanded study benchmarks 22 industry categories across four AI platforms during January through April 2026. The methodology tracks three metric categories: presence (citation frequency, mentions, cited pages, citation share), accuracy (sentiment and representation quality), and competitive positioning (citation share vs. rivals, competitor gaps, topic-level distribution).

The scale is significant. The dataset grew from 2,500 prompts to 126 million — a 50,400x expansion that moves AI visibility measurement from sample-based inference to population-level observation. At this scale, the study captures brand presence patterns that smaller prompt sets cannot detect: sector-level concentration, cross-platform consistency, and temporal stability.

Key findings from the dataset:

The 36-brand finding is the most structurally revealing. Among the brands that maintained cross-platform top-100 presence every month — YouTube, Google, Reddit, Amazon, Facebook, Apple, Walmart, Disney, Nintendo — the common trait is not content quality or optimization. It is platform scale and pre-existing brand recognition embedded in training data. This is visibility driven by brand mass, not by source authority on specific queries.

The Ghost Citation Problem: Why Visibility and Citation Authority Diverge #

Semrush's ghost citations study — a separate analysis of 3,981 domain appearances across 115 prompts in 14 countries and 4 AI engines — exposes the structural fault line between visibility and citation authority.

62% of all AI citations are ghost citations. In these cases, the AI engine used a domain as a source link but never named the brand in the response text. The reader sees the citation; the brand receives no recognition.

The engine-level breakdown shows why aggregate visibility metrics obscure source behavior:

The most striking finding: "there was almost no overlap between the brands ChatGPT cited and the brands Gemini named" for identical prompts. Two engines answering the same question selected different brands — and used different mechanisms (citation vs. naming) to surface them.

This divergence means a brand can have high "AI visibility" in aggregate while being invisible on the specific engine and query type that drives its buyer's research behavior. ChatGPT cites 87% of the time but names brands only 20.7% of the time. A domain earning ChatGPT citations may show strong visibility in a prompt-level tracker but zero brand recognition with the reader. Gemini reverses this: naming brands 83.7% of the time but citing only 21.4%. A brand that earns Gemini mentions may have strong naming recognition but no link-level citation authority.

Query phrasing compounds the divergence. Brand mention rates vary 30x to 50x depending on how the prompt is worded. Short conversational queries produce near-100% mention rates. Long structured queries produce 2-3% mention rates — on the same topic, about the same brand.

Visibility Measurement vs. Citation Authority Measurement #

The Semrush dataset and the Machine Relations Index measure related but structurally different properties of AI search behavior.

The distinction matters operationally. Semrush's visibility metric answers whether a brand appears — useful for brand awareness measurement. The MRI's citation authority metric answers why and where a domain gets cited as a source — useful for understanding what content properties make a domain citation-eligible and which engines select it.

Consider a concrete example. The MRI tracks Crunchbase's citation profile at the 99.7th percentile among 315 market databases, with 97 citations across 6 engines, 28 unique queries, and 9 verticals. The MRI can identify that Claude leads Crunchbase's citation share at 24.7% while ChatGPT accounts for only 3.1%. This engine-level granularity reveals that Crunchbase's citation authority is structurally concentrated in engines that value structured company data — a finding invisible in aggregate visibility scoring.

The Semrush approach would capture whether Crunchbase appears across 126 million prompts, yielding a visibility share metric. But without engine-level citation decomposition and source-role classification, the visibility share cannot distinguish between a domain that is cited because it is the best source for company data (Crunchbase) and a domain that is named because it is a well-known brand (Google, Amazon). Both register as "visible." Only one registers as citation-authoritative for specific query classes.

What the 615x Citation Variance Proves About Measurement #

Independent research from Superlines found citation rates vary 615x between AI platforms — comparing Grok's behavior to Claude's using March 2026 data. This variance aligns with MRI data showing that engine-level citation distribution varies dramatically by source role and domain.

Additional data points reinforce the measurement complexity:

• AI referral traffic grew 527% year over year through May 2025 (Previsible AI Traffic Report), with 87.4% driven by ChatGPT alone according to Conductor 2026 benchmarks.
• 93% of AI search sessions end without a website click. Organic CTR for queries with Google AI Overviews fell 61% — from 1.76% to 0.61% since mid-2024.
• AI traffic converts at 4.4x the rate of traditional organic visitors, per Semrush's own research.
• AI Overviews now appear in 25.11% of Google searches, up from 13.14% in March 2025.

The conversion premium and zero-click behavior create a paradox: AI citations drive higher-quality traffic, but 93% of AI interactions never produce a click. This means the citation itself — the domain's presence as a named, linked source in the AI response — is the primary brand touchpoint, not the click-through. Measurement that tracks click-level attribution misses 93% of the brand impact. Measurement that tracks prompt-level visibility captures presence but not the engine-specific citation mechanics that determine whether the brand is named, linked, or ghosted.

How Integrated Strategy Outperforms Siloed Approaches #

The Semrush Index reports that organizations unifying SEO and AI visibility achieved an 81% success rate for traffic and lead generation goals. Siloed approaches achieved 36% — a 2.25x gap.

Adobe CMO Rachel Thornton, quoted in the study: "Your AI narrative is becoming the decisive entry point" to customer experience.

This finding aligns with MRI-level evidence. Sources that earn high citation authority in the MRI share common structural properties: consistent page architecture, structured data with specific figures, and content that answers queries at multiple levels of specificity. These properties — documented in research on citation architecture — benefit both traditional search performance and AI citation eligibility. They are not separate optimization targets; they are the same structural investment producing returns in both channels.

The GEO market is projected to grow from $848 million in 2025 to $33.7 billion by 2034 at a 50.5% CAGR — indicating that the market has recognized the need for AI-specific measurement and optimization. But 54% of U.S. marketers plan GEO implementation within 3-6 months despite only 9% having comprehensive tracking tools. The execution-measurement gap mirrors the visibility-citation gap: practitioners are investing in optimization before they have established what they are optimizing for.

What Machine Relations Measurement Reveals That Visibility Cannot #

The gap between visibility and citation authority is not a product limitation. It is a structural difference in what each measurement paradigm captures.

Visibility measurement answers: "How often does my brand appear?" It is valuable for brand awareness tracking, competitive benchmarking across industries, and identifying which prompts trigger brand presence.

Citation authority measurement answers: "Why does this source get selected, by which engines, for which query types, and with what structural properties?" It is valuable for understanding the causal factors behind citation — and for identifying the content architecture changes that increase citation eligibility across engines.

The MRI tracks 6,160 domains across 6 AI engines, classifying each into source roles (market database, analyst research, editorial publisher, SaaS platform, etc.) and scoring across five components: engine breadth, query diversity, vertical spread, position quality, and temporal consistency. This decomposition reveals patterns invisible in aggregate visibility metrics:

• A domain can have Elite-tier citation authority in a narrow source role (e.g., Mordor Intelligence at the 98.9th percentile among market databases) while having minimal aggregate visibility across all prompts.
• Citation authority is role-specific. Forbes as an editorial publisher operates under entirely different citation dynamics than McKinsey as an analyst research source — even if both appear in the same AI response.
• Engine disagreement is structural, not random. MRI data shows that each engine's citation distribution for a given domain follows predictable patterns tied to the engine's retrieval architecture, not to brand visibility.

Semrush's 126-million-prompt dataset proves the scale of the AI visibility measurement market. The MRI's engine-level citation decomposition addresses the structural question that scale alone cannot answer: not whether your brand appears, but why AI engines select your domain as a source, and what you need to change to earn citation authority where it compounds.

FAQ #

What is the Semrush AI Visibility Index? #

A benchmarking study analyzing brand presence across ChatGPT, Gemini, Google AI Mode, and Google AI Overviews. The 2026 edition expanded from 2,500 prompts to 126 million U.S. prompts across 22 industries, tracking citation frequency, brand mentions, sentiment accuracy, and competitive positioning.

What are ghost citations in AI search? #

Ghost citations occur when an AI engine uses a domain as a source link but does not name the brand in the response text. Semrush's research found 62% of all AI citations are ghost citations. ChatGPT cites sources 87% of the time but names brands only 20.7%; Gemini reverses this pattern, naming brands 83.7% of the time but citing only 21.4%.

How does AI visibility differ from citation authority? #

Visibility measures whether a brand appears in AI responses across prompts. Citation authority measures why a specific domain is selected as a source — by which engines, for which queries, at what position, and with what structural properties. Visibility tracks brand presence; citation authority tracks the source-selection mechanics that determine how and why a domain earns citations. The Machine Relations Index measures citation authority across 6 AI engines using a five-component scoring methodology.

Why do different AI engines cite different brands for the same question? #

Each AI engine uses a different retrieval-augmented generation pipeline with distinct crawl scopes, source trust rankings, and response generation architectures. Semrush's ghost citations study found "almost no overlap between the brands ChatGPT cited and the brands Gemini named" for identical prompts. Research on AI citation divergence attributes this to structural differences in each engine's retrieval stack rather than to content quality differences.

Additional source context #

• Semrush Releases Expanded 2026 AI Visibility Index, Analyzing 126 Million AI Search Prompts | Morningstar Semrush Releases Expanded 2026 AI Visibility Index, Analyzing 126 Million AI Search Prompts The flagship study, scaled from an initial 2,500 prompts to 12 (Semrush Releases Expanded 2026 AI Visibility Index, Analyzing 126 Million AI Search Prompts | Morningstar (morningstar.c, 2026).
• Semrush Expands 2026 AI Visibility Index Study Sign Up We'll call you! (Semrush Expands 2026 AI Visibility Index Study (secureitworld.com), 2026).
• Semrush Features for AI Visibility ### Start Here ### Find the Right Tools ### Platform ### Top Apps Connect Semrush MCP to AI The world's most powerful traffic, visibility, and market dataset. (Semrush Features for AI Visibility (semrush.com), 2026).
• October 13, 2026London, UK AI Visibility Metrics # AI Visibility Metrics The AI Visibility Toolkit includes a set of unique metrics designed to help you measure brand visibility, track competitors, and understand AI-generated search performance. (AI Visibility Metrics (sr03.toolzpad.com), 2025).
• Where does the data in Semrush’s AI Visibility Toolkit come from? provides external context for Semrush AI visibility index 126 million prompts brand measurement.
• AI Visibility Index | Semrush for Enterprise provides external context for Semrush AI visibility index 126 million prompts brand measurement.