The First Large-Scale Index of AI Brand Visibility Is Here. The Findings Are Clear.

Similarweb's 2026 GenAI Brand Visibility Index tracked "brand mention share" across four AI platforms: ChatGPT, Gemini, Copilot, and Perplexity. The metric captures the percentage of AI-generated responses that name a specific brand for a given set of prompts. This is conceptually the same as what we call AI Share of Voice, the measure we have been tracking daily at Sill since launch.

The scope was significant: 113 brands across six sectors (Finance, Travel, Consumer Electronics, Beauty, Fashion, and News), with data from April 2025 through January 2026. In Finance alone, Similarweb analyzed 11,073 prompts. The index introduced three key outputs: a brand mention share ranking, a momentum score tracking change over time, and an "overachiever" metric identifying brands whose AI visibility far exceeds their traditional search presence.

The platform landscape underlying the index: ChatGPT holds roughly 79% of global generative AI web traffic. Gemini grew 157% between April and September 2025, reaching 1.1 billion monthly visits. Perplexity sits at 170 million monthly visits, Claude at 157 million. These are the surfaces where buyers increasingly discover brands.

The data reveals three patterns that hold across every sector: visibility concentrates at the top, momentum shifts can be dramatic, and content authority routinely outperforms brand scale.

The Electronics sector shows the most extreme concentration. Apple commands 54.38% brand mention share with a 94-point gap to tenth place. This mirrors what we found in our own analysis of 77 brands across 19 industries: AI visibility is a winner-take-most dynamic. The median AI Share of Voice score in our data was 3.8 out of 100. If you are not in the top few names an AI model associates with your category, you are effectively invisible.

The Fashion and News sectors reveal the most interesting momentum dynamics. Nike leads Fashion but is losing ground (-13.5 momentum), while New Balance and Uniqlo are climbing. In News, Reuters dominates AI visibility with 1.5 million monthly branded searches, while Fox News, with 42.1 million branded searches, ranks far lower. Traditional search presence does not predict AI visibility.

The most strategically significant metric in the index is the "overachiever" ranking, which measures how much higher a brand ranks in AI visibility compared to its branded search volume. This directly addresses the question we have been writing about since our research on where SEO ends and GEO begins: does traditional search dominance translate to AI visibility?

The answer, at scale, is no.

WhoWhatWear ranks 69 positions higher in AI visibility than its branded search volume would predict. NerdWallet and Bankrate outperform financial institutions with far larger marketing budgets. ScienceDirect, a research database, appears in AI news recommendations far above its consumer search footprint.

These brands share a common profile: deep, structured content with explicit attribute labeling and comprehensive comparison coverage. This aligns with the Ahrefs study of 75,000 brands, which found off-site web mention frequency correlates with AI visibility at 0.664, while backlinks correlate at just 0.218. It also aligns with the foundational GEO paper (Aggarwal et al., KDD 2024), which found that adding statistics to content delivers 30-40% visibility improvement, while keyword stuffing performs 10% worse than baseline.

The implication is clear: AI visibility rewards content authority over brand authority. A well-structured comparison guide with embedded statistics can outperform a brand with ten times the search volume if that brand's content is thin or gated.

The index contains a finding that will challenge how many marketers think about AI visibility ROI. AI platform visits grew 28.6% from January 2025 to January 2026. Over that same period, AI referrals to external websites stayed flat. LLMs contribute less than 1% of external web traffic.

More people are using AI platforms more often. They are not clicking through to external sites at a higher rate. This is not a failure of AI visibility. It is a structural feature of how AI discovery works. When a buyer asks ChatGPT "What CRM should I use for a 50-person team?" and gets a synthesized answer naming three products, many of those buyers go directly to the recommended product's website. In analytics, that shows up as direct traffic, not referral traffic.

There is a counterweight to this: ChatGPT visitors who do click through convert at 7% to transactional sites, compared to 5% from Google. Average session length is 15 minutes for ChatGPT referrals versus 8 minutes for Google, and pageviews per visit are 12 versus 9. The traffic is low volume but high quality.

This is exactly the attribution challenge we addressed in our piece on GEO attribution. Measuring AI visibility ROI through referral traffic alone understates the impact by an order of magnitude. The three-layer attribution model we proposed (simulated visibility, crawler verification, referral and lift attribution) accounts for the full picture, including the branded search lift and direct traffic that AI recommendations generate.

The Similarweb data, combined with the independent research we have been tracking, converges on a consistent set of content signals that AI platforms reward.

The paywall finding deserves particular attention. In the News sector, the New York Times and Wall Street Journal rank 8th and 9th despite being among the most recognized news brands in the world. Reuters, which keeps most content openly accessible, leads the sector. AI models cannot cite content they cannot retrieve. This is a direct consequence of the chat-layer retrieval architecture we described: the chat interface retrieves live web content, and paywalled content is excluded from that retrieval.

CeraVe's dominance in Beauty follows the same logic. CeraVe's product pages feature explicit ingredient lists, structured comparisons, and detailed attribute labeling. It has deep presence across YouTube skincare channels and Reddit communities like r/SkincareAddiction. Ulta, despite 10x the branded search volume, relies on a marketplace model with thinner product-level content. AI models reward the depth, not the brand size.

The index tracks momentum over a ten-month window (April 2025 through January 2026). The swings are large. Ulta Beauty surged to a 319.0 momentum index (up 219 points). B&H Photo reached 296.9 (up 196.9 points). The Washington Post climbed to 271.5. On the other side, the Wall Street Journal fell to 52.3 (down 47.7 points), and Nike declined by 13.5 points.

These momentum shifts are consistent with the volatility we see in our own daily monitoring. Research has shown that AI Overview content changes roughly 70% of the time for identical queries, and nearly half of cited sources get replaced between runs. A brand can move from invisible to recommended, or vice versa, within weeks. A quarterly or monthly check is too slow.

This is why Sill runs daily monitoring across six platforms. A single snapshot gives you a score. Daily tracking shows you the trajectory. When a competitor launches a content campaign or a major publication mentions your brand, you see the impact on your Share of Voice within days, not months.

The Similarweb index measured visibility across four platforms simultaneously. One of the report's structural insights is that brands appearing consistently across all four platforms have more durable competitive positioning than brands concentrated on a single platform.

This connects directly to what we have measured. Only 11% of domains cited by ChatGPT are also cited by Perplexity. Each platform has different retrieval pipelines: ChatGPT draws on Bing search, Perplexity aggressively retrieves from Reddit and YouTube, Gemini pulls from Google Search, and Copilot integrates Bing with enterprise context. CMU's AutoGEO research (Wu et al., 2025) found that engine-specific optimization rules outperform generic strategies, achieving 35.99% average improvement.

Sill monitors ChatGPT, Perplexity, Gemini, Google AI Overviews, Copilot, and Grok independently. Each platform gets its own SOV score, sentiment analysis, and citation tracking. When your team makes a content change, you can see whether it moves the needle on Perplexity but not ChatGPT, or improves Gemini visibility while Copilot stays flat. That platform-level granularity is essential for a targeted GEO strategy.

The Similarweb index validates what we have been building toward at Sill. AI visibility is a measurable, trackable channel. It operates by different rules than traditional search. And brands that understand those rules early are building a compounding advantage.

Specifically, the index confirms four principles we have been tracking across our own research:

1. Domain authority does not predict AI visibility. The SearchAtlas study of 21,767 domains found correlations of r = -0.12 for ChatGPT and -0.18 for Perplexity. The Similarweb data shows the same pattern: NerdWallet and Bankrate outperform financial institutions, CeraVe outperforms Ulta, Reuters outperforms Fox News.
2. Content depth and structure matter more than brand scale. The overachiever brands share common traits: comprehensive comparison content, explicit attribute labeling, embedded statistics, and open access for AI crawlers.
3. Off-site presence is the strongest controllable lever. The Ahrefs study found a 0.664 correlation between web mention frequency and AI visibility. YouTube accounts for 29.5% of all AI citations. Reddit appears in 46.7% of Perplexity citations. Brands winning in AI are present across the UGC and publisher ecosystem.
4. AI visibility changes fast. Momentum swings of 200+ points over ten months are common. Daily monitoring is the minimum viable measurement cadence.

Consumer adoption data reinforces the urgency. Similarweb reports that 35% of US consumers now use AI for discovery (compared to 13.6% for search), 32.9% for evaluation (compared to 15% for search), and 24.3% for final purchase decisions (compared to 22.1% for search). AI holds a 2:1 or greater advantage over traditional search through every funnel stage except the final transaction. The G2 Buyer Behavior Report found that GenAI chatbots are already the number-one source influencing B2B vendor shortlists.

The Similarweb index provides a valuable macro view of brand visibility by sector. At the individual brand level, acting on AI visibility requires deeper measurement. Sill provides three layers that go beyond brand mention rankings.

Daily Share of Voice tracking across ChatGPT, Perplexity, Gemini, Google AI Overviews, Copilot, and Grok. Unlike the Similarweb index, which captures a snapshot, Sill monitors daily through the actual chat interfaces with web search enabled and citations included. This means every data point reflects what a real buyer would see when asking AI for a recommendation.

Semantic positioning via the Semantic Map. The Similarweb index tells you whether you are mentioned. It does not tell you how AI perceives your brand relative to competitors on dimensions that matter to buyers. Sill's Semantic Map plots brands on configurable axes (price vs. innovation, trust vs. specialization, or any dimension you define), showing perception gaps that no visibility score alone reveals.

Actionable diagnostics and GEO recommendations. Across 748 GEO recommendations generated for 62 brands, 87% were on-site fixes brands can implement immediately: schema markup, answer capsules, embedded statistics, comparison content. Sill reports score your pages against 18+ peer-reviewed GEO factors and prioritize what to fix first.