By Ammar Sabilarrohman 
As generative artificial intelligence continues to reshape the digital landscape, the mechanisms of online visibility are undergoing a profound paradigm shift. For over two decades, search engine optimization (SEO) was governed by predictable ranking factors: backlinks, keyword density, technical performance, and domain authority. However, the rise of Large Language Models (LLMs) as primary information-retrieval engines has introduced a highly volatile, conversational ecosystem where traditional rules no longer guarantee visibility.
A groundbreaking study conducted by search marketing intelligence platform Semrush, in collaboration with SEO strategist Kevin Indig, has pulled back the curtain on how brands are represented within ChatGPT. The comprehensive analysis reveals that the generative search landscape remains largely unstructured and highly competitive: only 15.2% of analyzed ChatGPT topic categories possess a clear brand "owner" across related buyer questions.
For enterprise brands and digital marketers, this statistic is a double-edged sword. It indicates that approximately 85% of product and service categories have no dominant brand presence within ChatGPT’s conversational outputs. While this represents a massive, untapped opportunity for agile brands to establish early dominance in AI-driven search, it also highlights the precarious nature of visibility in the age of Generative Engine Optimization (GEO).
Main Facts: The New Rules of AI Brand Visibility
The joint study by Semrush and Kevin Indig analyzed how ChatGPT handles queries across a diverse spectrum of commercial topics. The core findings challenge several long-held assumptions about digital marketing, brand authority, and search engine behaviors.
Defining "Category Ownership" in the Age of LLMs
In traditional search, "ownership" of a category is often defined by ranking in the top three organic spots on Google for high-volume keywords. In the conversational paradigm of ChatGPT, Semrush had to establish a new, multi-layered definition for "category ownership."
To be classified as a category owner, a brand had to meet three stringent criteria:
- High Share of Mentions: The brand must secure the highest volume of mentions within the generated text.
- Consistency Across Prompts: The brand must appear in at least four out of five related buyer prompts within a specific topic category.
- A Clear Competitive Margin: The brand must lead its closest competitor (the runner-up) by at least 5 percentage points in total share of mentions.
Under these rigorous criteria, only 15.2% of the analyzed categories had a definitive leader. In the remaining 84.8% of categories, ChatGPT did not consistently favor any single brand, presenting a highly fragmented landscape where different brands appeared depending on how a question was framed.
+-----------------------------------------------------------------------+
| ChatGPT Category Ownership |
+-----------------------------------------------------------------------+
| [███████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░] |
| |
| ■ Clear Brand Owner: 15.2% |
| ■ No Dominant Brand: 84.8% |
+-----------------------------------------------------------------------+The Disconnect Between Mentions and Citations
One of the study’s most critical revelations is that being cited is not the same as being recommended.
Semrush measured category ownership based on brand mentions within ChatGPT’s natural language answer text, rather than the external links and citations appended at the bottom of the response. The data revealed that the most frequently cited domains were rarely the most frequently mentioned brands in the conversational narrative. This suggests that ChatGPT often pulls factual data or links from one set of web sources (citations) while synthesizing its brand recommendations and comparisons from a different layer of its parametric memory or retrieval models.
The Limits of Traditional SEO Metrics
Perhaps the most alarming finding for legacy brands is that traditional SEO strength has limited predictive value in ChatGPT. High domain authority, massive backlink profiles, and top-tier organic rankings on Google do not automatically translate to dominance in LLM responses. A brand can dominate Google’s Search Engine Results Pages (SERPs) for a specific product category but remain virtually invisible when ChatGPT synthesizes a list of recommendations for a user.
Chronology: Volatility and the Evolution of AI Authority
The study was structured as a longitudinal analysis, tracking ChatGPT’s behavior over a six-month period from January through June 2026. By monitoring the data month-over-month, researchers were able to map out how brand associations evolve within OpenAI’s model and evaluate the stability of AI search visibility over time.
2026 Longitudinal Study Timeline
+-----------------------------------------+
| JAN | FEB | MAR | APR | MAY | JUN |
+-----------------------------------------+
_________________________________________/
|
Analysis of 1,094 US Categories
Monitoring 50,000+ Brands over timeThe Stability of Established Leaders
The chronological data revealed a stark contrast between established category leaders and those in emerging or contested spaces.
For the 15.2% of brands that managed to establish a clear category ownership, their position was remarkably secure. Across month-over-month comparisons, clear category owners maintained their first-place standing in 90.4% of instances. Once ChatGPT’s training data, fine-tuning, or retrieval-augmented generation (RAG) pathways solidify a brand as the definitive answer for a category, that association becomes highly resistant to change.
The Volatility of the Unsettled Long Tail
Conversely, in categories where no clear brand owner existed, or where the leader’s margin was slim, market volatility was exceptionally high.
In emerging and unsettled categories, the top brand switched frequently. Across 5,470 month-over-month comparisons of these contested categories, the leading brand changed in 1,950 instances (approximately 35.6% of the time). This indicates that without a strong, mathematically significant lead, brand visibility in ChatGPT is highly susceptible to minor algorithmic updates, shifts in web crawling patterns, or changes in user prompting behavior.
Supporting Data: A Deep Dive into the Numbers
The scale of the Semrush and Kevin Indig study provides one of the most robust datasets on generative AI search behavior compiled to date.
Dataset Metrics
To ensure the statistical relevance of the study, Semrush utilized its AI Visibility Toolkit to analyze a massive cross-section of the digital economy:
- Categories Analyzed: 1,094 distinct U.S. product and service categories.
- Brands Monitored: Over 50,000 unique brands.
- Domains Tracked: 220,000 domains.
- Citations Evaluated: 600,000 citations.
- URLs Analyzed: 220,000 source URLs.
The Five-Prompt Buyer’s Journey Framework
To simulate real-world consumer behavior, Semrush designed five specific prompts for each of the 1,094 categories. These prompts mapped directly to key stages of the modern marketing funnel:
- Definitions: Establishing foundational knowledge (e.g., "What is enterprise CRM software?")
- Comparisons: Evaluating competing options (e.g., "Compare HubSpot and Salesforce CRM.")
- Alternatives: Finding substitute goods (e.g., "What are alternatives to Monday.com?")
- Use Cases: Solving specific problems (e.g., "What is the best CRM for small real estate agencies?")
- Buying Decisions: Intent-driven transactional queries (e.g., "What is the best enterprise CRM to buy in 2026?")
The study tracked whether a brand could maintain its presence across all five of these distinct prompts within its category. The fact that only 15.2% of brands could do so highlights the difficulty of maintaining a cohesive presence across the entire consumer decision-making journey in an AI environment.
High-Demand vs. Low-Demand Categories
The researchers split the analyzed categories into two equal halves based on AI search demand (how frequently users asked ChatGPT questions about these topics). The results showed an inverse relationship between search popularity and brand dominance:
Clear Brand Ownership by Category Demand:
High-Demand Categories (Top 50%)
[██████████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░] 11.3%
Low-Demand Categories (Bottom 50%)
[███████████████████░░░░░░░░░░░░░░░░░░░░░░░░░░] 19.0%- High-Demand Categories: Only 11.3% of topics in the top half of search demand had a clear brand owner.
- Low-Demand Categories: In contrast, 19.0% of topics in the bottom half had a clear owner.
This discrepancy suggests that high-volume, highly lucrative categories are characterized by intense commercial noise, diverse web sources, and conflicting information, making it difficult for ChatGPT to settle on a single, dominant brand. In contrast, niche or low-demand categories have fewer dominant sources of truth, allowing specific brands to establish a monopoly on the model’s output more easily.
The Margin of Dominance
The study also analyzed the exact tipping points where a brand’s leadership became unstable. The margin of a brand’s lead over its runner-up was highly indicative of its future retention of that spot:
- The Vulnerability Threshold: When a leading brand eventually lost its first-place ranking in subsequent months, its historical lead margin had averaged just 1.3 percentage points.
- The Retention Threshold: When a brand successfully defended its first-place position, its typical lead margin was 2.9 percentage points—more than double the margin of vulnerable leaders.
Official Responses and Industry Commentary
The implications of the study have reverberated across the search marketing community, prompting responses from digital strategists, SEO agencies, and the study’s authors.
Kevin Indig on the Limitations of Traditional SEO
Kevin Indig, founder of the industry publication Growth Memo and co-author of the study, emphasized that marketers must abandon the comfort of legacy metrics when optimizing for AI engines:
"Traditional SEO metrics aren’t enough to explain who owns a topic. While they play their role, there’s more to it. LLMs operate on semantic associations, context, and entity relationships rather than simple keyword matching and link equity. A domain with millions of backlinks might be completely ignored by ChatGPT if its content fails to directly answer the specific semantic intent of the user’s prompt, or if the model’s training data associates a competitor more strongly with the underlying concept."
Semrush on the Reality of the "Zero-Click" Era
In a statement accompanying the study’s release, Semrush highlighted that the rise of generative AI search engines represents a transition into a deeper "zero-click" environment, where users obtain direct recommendations without ever visiting a brand’s website:
"If AI can’t find you, customers won’t either. The battleground for visibility has moved from the search results page to the LLM response itself. Tracking your brand’s visibility across AI search, uncovering missed opportunities, and growing your presence where customers are asking natural-language questions is no longer a forward-looking strategy—it is an immediate survival requirement for digital businesses."
The Editorial Context
It is worth noting that Search Engine Land, which published the initial coverage of this study, is owned by Semrush. However, the publication maintains its commitment to editorial independence, and the raw scale of the data—spanning 50,000 brands and over half a million citations—has been widely accepted by the broader marketing community as a landmark reference point for Generative Engine Optimization.
Implications: The Playbook for Generative Engine Optimization (GEO)
The findings of the Semrush study have profound implications for CMOs, brand managers, and digital marketers. The fact that 85% of categories lack a dominant brand means that the generative AI space is currently an "open season."
To capture market share in this new environment, organizations must pivot from traditional SEO to Generative Engine Optimization (GEO).
+-------------------------------------------------------------------+
| Traditional SEO vs. Generative Engine (GEO) |
+-------------------------------------------------------------------+
| Metric / Strategy | Traditional SEO | Generative SEO (GEO) |
|-------------------|----------------------|------------------------|
| Core Focus | Keywords & Backlinks | Entities & Context |
| Success Metric | SERP Position 1-10 | Share of Voice (SOV) |
| Primary Goal | Drive Website Clicks | Direct Brand Mention |
| Target Engine | Indexing Crawlers | LLM Parametric Memory |
+-------------------------------------------------------------------+1. Optimize for the Full Buyer’s Journey
Because category ownership requires a brand to appear in at least four out of five related prompts, marketers can no longer rely on single-keyword optimization. Content strategies must be structured to answer the entire conversational funnel.
- Brands must create content that defines their category, compares them objectively to competitors, positions them as the premier alternative to legacy players, highlights specific niche use cases, and directly addresses the final buying decision.
2. Focus on Natural Language Mentions, Not Just Links
Since ChatGPT frequently mentions brands in its generated text without necessarily linking to them in the citation footer, digital PR and brand sentiment have become critical.
- To be mentioned by an LLM, a brand must exist in the model’s training data and RAG retrieval pipelines as a highly trusted entity. This requires securing high-authority coverage in editorial publications, active discussions on forums like Reddit and Quora (which are heavily crawled by AI developers), and consistent positive reviews on trusted third-party platforms.
3. Establish a Wide Margin of Safety
With the data showing that a lead margin of 1.3 percentage points is highly unstable, brands cannot afford to be complacent with a narrow lead in AI visibility.
- Once a brand gains a foothold in ChatGPT’s recommendations, it must aggressively expand its content footprint to push its lead past the 2.9 percentage point "safety threshold." This involves saturating the digital ecosystem with semantic variations of their core product offerings, ensuring that no matter how a user phrases a prompt, the model’s probabilistic calculations point back to their brand.
4. Leverage Niche Categories First
Given that low-demand categories exhibit nearly double the rate of clear brand ownership (19%) compared to high-demand categories (11.3%), brands should consider a "bottom-up" approach to GEO.
- Instead of trying to dominate highly competitive, broad category terms immediately, companies should focus on dominating niche, long-tail, and industry-specific prompts. Once dominance is established in these lower-volume categories, the resulting semantic associations can help the brand climb into broader, higher-demand categories.
Conclusion: The Cost of Inaction
The Semrush study makes one thing clear: the window of opportunity to define category leadership in generative AI is open, but it will not remain so indefinitely. The brands that successfully establish a clear lead of over 3% in their respective categories today will enjoy a 90.4% retention rate, effectively locking out their competitors for years to come. For the remaining 85% of categories, the race is on—and the rules of engagement have been rewritten.
Reynand Wu 
Azzam Bilal Chamdy 
Evan Lee Salim 