Executive Overview: A Landscape in Flux
The relationship between Generative AI (GenAI) and traditional search engines has reached a point of profound complexity. According to the latest "State of GenAI" report released by Graphite, an agency specializing in SEO and AEO (Answer Engine Optimization), global engagement with six leading AI assistants reached 76 billion monthly sessions in September 2026. This represents a staggering 37% year-over-year increase, signaling that AI adoption is far from cooling down.
However, the report serves as a cautionary tale for data analysts and digital marketers alike. In a significant shift from its March edition, Graphite has moved away from direct comparisons between AI sessions and traditional search engine traffic. The reason? A newfound humility regarding the limitations of web analytics. The agency admits that previous models likely undercounted Google’s massive footprint—specifically via the Android app—rendering the historical “AI vs. Search” ratios unreliable. As the lines between search, translation, image recognition, and AI-driven discovery blur, the industry is discovering that measuring the "AI revolution" is significantly harder than anticipated.
Detailed Chronology: The Evolution of Graphite’s Methodology
To understand the current state of the industry, one must look at how the data has evolved over the past year. In March 2026, the industry was captivated by reports suggesting that AI sessions were climbing to roughly 56% the size of global search engine traffic. This provided a compelling, albeit potentially flawed, narrative of a "search revolution."
The "Android" Realization
The discrepancy lies in the definition of a "search." Graphite’s March report relied on web-based traffic measurements for search engines, largely because mobile app usage for search was historically negligible. However, the November report highlights a critical blind spot: the Google Android app.
With an estimated 11.8 billion monthly sessions in the U.S. alone, the Google app serves as a Swiss Army knife for modern users, facilitating image recognition, translation, and the Google Discover feed alongside traditional keyword searches. Graphite now concedes that it lacks a reliable data source to parse what portion of these 11.8 billion sessions constitutes "general search." Consequently, the agency has effectively retracted its ability to benchmark AI against search, noting that it "can no longer reliably compare" the two mediums.
This admission creates a ripple effect throughout the industry. The March report’s assertion that ChatGPT accounted for 20% of global search-related traffic was predicated on these web-only counts. While that page remains live—adorned with a link to the new, more cautious report—the industry is being forced to grapple with the reality that early, headline-grabbing metrics were likely based on incomplete datasets.
Supporting Context & Metrics: The Mechanics of Growth
Despite the methodological challenges, the underlying growth trends in the Graphite report provide a robust picture of how users are interacting with these new tools.
The Plateau and the Surge
The data reveals a fascinating narrative of adoption. Between July and December 2025, global activity with the six major AI assistants stagnated, experiencing a modest 2.3% growth—a period some dubbed a "plateau." This phase proved to be the calm before the storm. From January through September 2026, global activity surged by 19%. In the United States, the momentum is even more pronounced, with year-over-year session activity up 45% as of September.
Platform Performance: The Giants and the Challengers
- ChatGPT: Maintaining its position as the market leader, ChatGPT saw 24% year-over-year growth globally. It continues to act as the primary engine of industry expansion, accounting for 58.9% of all new sessions added during the tracking period.
- Claude: The most explosive growth came from Anthropic’s Claude, which saw a staggering 1,901% increase in global sessions. While starting from a smaller base, Claude contributed 26% of the additional sessions, signaling a diversification in the AI landscape.
- Perplexity: Notably, Perplexity was the only major assistant in the study to show a decline, dropping 27% globally. This suggests that as the market matures, users are becoming more selective, perhaps favoring the multimodal capabilities of larger ecosystems like ChatGPT and Gemini over search-focused AI interfaces.
The Mobile Dominance
The era of the "desktop-first" web is effectively over for the AI sector. In the third quarter of 2026, 89% of global AI sessions occurred on mobile devices (via app or web). In the U.S., mobile accounted for 78%. This mobile-first behavior is precisely why the exclusion of app-based data in earlier reports led to such significant analytical errors.
AI Referral Traffic: The "Less Than 1%" Reality
While session counts are rising, the direct impact on website traffic remains surprisingly modest. Graphite reports that AI accounts for less than 1% of total traffic to websites on average.
To put this into perspective, a chart of U.S. traffic from the second quarter of 2026 shows that organic search still drives 49 times the volume of AI referrals, while paid search drives eight times. Despite the massive noise surrounding "AI replacing search," the tangible flow of users to websites via AI chat interfaces remains a niche channel.
However, growth is occurring at a rapid clip. AI referral traffic grew 92% year-over-year worldwide as of July, and 75% in the U.S. The breakdown of who is driving this traffic remains consistent with the session data:
- ChatGPT: 60% global share of AI referral traffic.
- Gemini: 14% global share.
- Claude: 10% global share.
In the U.S. market, ChatGPT’s dominance is even more pronounced at 66%, with Claude (11%) and Gemini (10%) trailing closely.
Official Statements and Industry Friction
The lack of standard metrics remains a significant hurdle for businesses trying to allocate their marketing budgets. OpenAI recently claimed that ChatGPT receives approximately one-fifth as many search-like queries as Google. However, this claim is riddled with definitional ambiguity. OpenAI counts a "search" every time a query triggers a web search action, whereas analysts like Graphite count total website visits and app sessions.
When comparing these figures, the industry is effectively trying to measure the volume of two different fluids with two different types of volumetric flasks. As the report wisely cautions: "Before an AI-to-search ratio goes into a channel plan, check how its source counted Google."
The fragmentation of data sources—where Similarweb provides the raw traffic data and agencies like Graphite interpret it through an SEO lens—means that every report carries the inherent biases and limitations of its data provider.
Future Outlook: Beyond the Click
As we look toward the future, the primary challenge for the digital marketing industry is the "dark traffic" problem. Graphite cites anecdotal evidence suggesting that many visits prompted by AI answers are being credited to "direct" or "branded search" traffic. Because users are increasingly interacting with brands inside the AI interface—asking questions about products, services, or prices—the traditional "click" is becoming less common.
The Branded Search Connection
This aligns with previous data from the first half of 2026, which indicated that when ChatGPT recommends a brand, a significant portion of the subsequent site traffic arrives via a branded search query (a user typing the brand name into Google). If a consumer asks ChatGPT, "What is the best CRM for my business?" and the AI suggests a brand, the user is likely to then search for that brand by name to visit the site. In this scenario, the AI acts as a discovery layer, while the search engine acts as the navigator—a symbiosis that current attribution models are ill-equipped to track.
The Road Ahead
The "State of GenAI" report serves as a timely reminder that the AI revolution is not a replacement of the web, but an evolution of it. The "plateau" seen in late 2025 suggests that the initial novelty of LLMs has worn off, replaced by a period of utilitarian, habit-forming use.
For businesses, the lesson is clear: relying on singular, high-level metrics like "AI vs. Search" is a fool’s errand. The future of digital strategy lies in understanding the nuance of the user journey—a journey that now spans across app-based AI discovery, branded search navigation, and direct web engagement. As measurement tools catch up to user behavior, the industry will likely find that AI is not destroying the search ecosystem, but rather fundamentally rewriting the rules of how users find and interact with the brands they trust.