There is a number that has been circulating in digital marketing conversations over the past several months, and it deserves more careful examination than it has received: AI assistants now generate sessions equivalent to 56% of global search engine volume. That figure comes from a study conducted by Ethan Smith, CEO of Graphite.io, and while it sounds extraordinary, the methodology behind it is arguably more important than the headline itself.
The reason the number is so large — and the reason most industry estimates have dramatically undercounted AI usage — comes down to a single, widely overlooked factor: mobile apps. Almost every comparison you have read between ChatGPT and Google has measured website traffic. And almost every such comparison has, in doing so, missed between 75% and 83% of all AI activity happening on the planet right now.
If you are making decisions about content strategy, search optimization, media spend, or brand visibility in 2025 and beyond, this study changes the framework. Not because it signals the death of Google — the data does not support that — but because it reveals that the total landscape of information retrieval has grown significantly, and that a large, fast-growing portion of that landscape has been effectively invisible to most marketing analysts.
This post unpacks the data layer by layer, examines what it means for your search strategy, and gives you a practical framework for operating in a world where both traditional search engines and AI assistants require deliberate optimization.
The Study: What Was Actually Measured
The Graphite.io analysis is notable for what it included, not just what it found. Most prior comparisons of AI tools versus search engines took a simple approach: count website visits to Google.com, count website visits to ChatGPT.com, and compare the two numbers. That approach has the virtue of simplicity. It also has the flaw of being structurally incomplete.
The Graphite.io methodology was different in three meaningful ways.
First, it combined both web traffic and mobile app usage across the platforms being analyzed. This matters enormously because the majority of ChatGPT, Gemini, Perplexity, Grok, and Claude usage happens through native iOS and Android applications, not through a browser visiting a .com domain. Web analytics tools like Similarweb or SemRush can measure website visits reasonably well. They cannot directly measure what happens inside a mobile app.
Second, it analyzed the five largest AI products together, rather than treating ChatGPT as a proxy for all AI usage. ChatGPT does dominate — it represents approximately 89% of total global AI sessions across the platforms measured — but Gemini, Perplexity, Grok, and Claude together constitute a meaningful portion of the remaining 11%.
Third, it compared these AI platforms against the six largest search engines, not just Google. This gives a more accurate picture of the overall search ecosystem — one that includes Bing, Yahoo, Baidu, Yandex, and DuckDuckGo alongside Google’s dominant position.
The result of this methodology is a dataset that, according to the study, is 4 to 5 times larger than what most web-traffic-only comparisons would produce. That is not a rounding error. It is a structural gap in how the industry has been thinking about AI usage.
The 45 Billion Session Reality
The core finding of the Graphite.io study is this: AI platforms collectively generate 45 billion monthly sessions worldwide.
To put that figure in context, traditional search engines process an estimated 8 to 9 trillion queries per year, which translates to roughly 700 to 750 billion monthly sessions when aggregated across all major engines. Forty-five billion AI sessions against that baseline equals approximately 56% of global search engine volume — not 56% of Google’s volume specifically, but 56% of the entire search engine landscape combined.
In the United States, the picture is different. AI accounts for approximately 5.4 billion monthly sessions in the U.S., which the study calculates as 34% of U.S. search engine volume. The lower U.S. percentage relative to the global figure reflects two dynamics: the United States has very high traditional search activity per capita, and international markets (particularly in Asia) are adopting AI tools at rates that outpace the U.S. on a proportional basis.
These numbers deserve neither panic nor dismissal. They should be read as evidence of market expansion. The study calculates that total usage across both search engines and AI assistants has grown 26% globally since 2023. AI is not cannibalizing search in a simple, direct substitution. It is expanding the total volume of information-seeking activity. People are doing both — they are still searching on Google, and they are also asking questions of AI assistants. The two behaviors overlap, but they are not identical.
The Mobile App Gap: Why Previous Estimates Were Off by 4–5x
This is the most technically important finding in the entire study, and it is worth dwelling on.
When analysts reported that ChatGPT had “3 billion monthly visits” in late 2024, those figures came from web traffic measurement tools that counted visits to chat.openai.com or chatgpt.com. Those tools are excellent at what they do. What they cannot do is count what happens inside the ChatGPT app on your iPhone or Android device. Every question you ask ChatGPT through the app is invisible to web analytics platforms.
The Graphite.io data estimates that 83% of global AI usage occurs inside mobile apps. In the United States, that figure is 75% — slightly lower globally but still the overwhelming majority. This means that for every session tracked by web analytics tools, approximately four to five additional sessions were happening in mobile environments that those tools simply could not see.
This has significant implications for how the industry has framed the AI-versus-search debate. If you have seen charts showing ChatGPT with a few billion monthly visits and Google with 100+ billion, and concluded that AI is still a marginal player in the information retrieval landscape, you were working with data that represented perhaps 20% of the actual AI usage picture.
The mobile app gap is not a flaw in the existing tools — it is a structural limitation of web traffic measurement applied to a landscape that has moved significantly beyond the web browser. Marketers who relied on Similarweb or comparable tools to benchmark AI adoption levels may need to revisit their assumptions.
For context: ChatGPT became the most downloaded app globally in 2025, surpassing TikTok. Its session duration on mobile is also significantly longer than that of traditional search engines. When users engage with ChatGPT, they are spending an average of 13+ minutes per session compared to just over 6 minutes on Google. The nature of the interaction is fundamentally different — deeper, more conversational, and more likely to result in a decision without requiring a follow-up search.
ChatGPT’s Dominance and the AI Landscape Breakdown
Within the AI assistant space itself, the concentration is striking. ChatGPT represents approximately 89% of total global AI sessions across the five major platforms analyzed. That is a level of platform concentration that even Google, at the height of its search dominance, rarely achieved in a given category.
What this means practically is that, when discussing AI search optimization, ChatGPT is not one of several equally important targets. It is the primary target, with Gemini, Perplexity, Grok, and Claude representing a collectively meaningful but individually smaller set of secondary channels.
Each of these secondary platforms serves a different use pattern:
Gemini is deeply embedded in the Google ecosystem — within Google Search itself (as AI Overviews), in Android devices, and across Google Workspace. Much of Gemini’s usage is invisible as discrete AI activity because it surfaces inside what still looks like a Google search experience. This integration makes Gemini harder to measure but potentially more influential than its standalone traffic numbers suggest.
Perplexity has built a reputation as the AI platform that cites its sources most transparently. For publishers and content creators, this is particularly important — Perplexity’s outbound referral rate is higher than ChatGPT’s, meaning a mention in a Perplexity answer is more likely to generate an actual click to your website. Its user base, estimated at well over 100 million monthly users by late 2025, skews toward researchers, analysts, and knowledge workers.
Claude, developed by Anthropic, has carved out a professional and academic niche. Its outbound links tend to lead to search engines and research institutions, reflecting a user base engaged in deep, multi-step research tasks rather than quick informational lookups.
Grok, integrated with the X platform, brings a real-time social media intelligence angle that the other assistants lack. Its ability to surface information from ongoing conversations and breaking news on X gives it a distinct role in the information ecosystem, particularly for trend-sensitive queries.
Despite their differences, all five platforms share a common trait that distinguishes them from traditional search engines: they synthesize answers rather than returning lists of links. This changes the nature of what it means to “rank” on these platforms. Visibility in an AI assistant’s answer is not about holding a top-10 position in a results page — it is about being cited, referenced, or incorporated into a synthesized response that the user receives as a single, authoritative answer.
Google’s Market Share Is Eroding, But Context Matters
One of the most quoted statistics from the study is that Google’s share of search-related activity fell from 89% in 2023 to approximately 71% in Q4 2025. That is an 18-percentage-point decline in roughly two years. It sounds dramatic, and it is a real trend. But several important caveats apply.
First, these figures are specific to the study’s framing — they measure Google’s share of combined search engine plus AI assistant sessions, not Google’s share of search alone. Within the traditional search engine category, Google still commands an overwhelming majority. StatCounter data from 2025 shows Google holding roughly 86–90% of global search market share, depending on the measurement period. The erosion relative to AI is real; the collapse of Google’s search dominance is not.
Second, Google is not a passive bystander in this shift. AI Overviews — Google’s integrated AI-generated summaries that appear at the top of search results — reached over 1.5 billion monthly users by Q1 2025. Google’s own Gemini model is the engine behind these Overviews. In other words, some portion of the sessions being attributed to “AI” in aggregate analyses is actually Google-mediated AI, just in a new form factor.
Third, the revenue picture is not as alarming as market share numbers might suggest. Google’s transactional query share — the searches where users intend to buy something, book a service, or make a commercial decision — remains near 90%. These are the queries with the highest advertising value. Users buying car insurance or comparing SaaS platforms are still predominantly doing so through Google. The queries migrating to AI assistants tend to be informational in nature.
That said, Gartner’s prediction that traditional search engine volume will decline 25% by 2026 should be taken seriously. Even if transactional queries stay on Google, a 25% reduction in informational query traffic has meaningful implications for content publishers, media sites, educational platforms, and any brand that relies on informational content to attract customers at the top of the funnel.
The “Search-Like” Prompt Distinction: 28% vs. 56%
The 56% figure is the headline number from the Graphite.io study, but a more nuanced figure is arguably more useful for marketers: when you isolate only “search-like” prompts — those where a user is asking a question or seeking information — AI activity equals 28% of global search volume and 17% in the United States.
The distinction the study makes is between three types of AI usage:
- “Asking” — information-seeking queries, the closest behavioral equivalent to a search engine query
- “Doing” — task completion, such as drafting an email, writing code, or editing a document
- “Expressing” — creative or generative tasks like writing poetry, generating images, or roleplaying
The 56% headline figure includes all three categories. The 28% figure isolates only “asking” behavior. OpenAI’s own research estimates that approximately 52% of all prompts submitted to ChatGPT are information-seeking in nature, meaning they are functionally similar to what a person would previously have typed into Google.
This 28% global / 17% U.S. figure is the more directly relevant number for SEO professionals and search marketers. It represents the share of information-seeking activity that has migrated to AI platforms — and it suggests that while AI has captured a significant and growing portion of informational queries, the majority of search-intent behavior is still happening on traditional search engines.
The practical implication: your content strategy needs to serve both environments. Content optimized only for traditional search rankings is increasingly missing a population of information-seekers who will never see it. Content optimized only for AI citation lacks the structured signals needed to attract the still-dominant traditional search audience.
U.S. vs. Global: Two Very Different Stories
The U.S. numbers in this study deserve particular attention because they reveal a pattern of behavior that is likely to spread to other markets over time.
Globally, AI usage appears to have plateaued since approximately July 2025. New user growth has slowed in markets like Europe and parts of Asia, and session volume has stabilized in those regions.
The United States, however, is experiencing a very different trajectory. U.S. AI usage was up approximately 300% year-over-year by December 2025. This divergence suggests that the U.S. market is in an earlier, faster-growth phase of AI adoption relative to the global curve, or that specific cultural, linguistic, and product-accessibility factors are accelerating U.S. uptake.
This has direct implications for U.S.-focused brands. If AI usage is growing at 300% annually in the U.S., and 37% of U.S. consumers are already beginning their information journeys with AI tools rather than traditional search engines (per a separate Eight Oh Two Marketing study of 2026 AI + Search Behavior), then the window for establishing AI visibility is open now. Brands that build presence in AI-generated answers while the landscape is still forming will face less competition than those who wait until AI optimization becomes as commoditized as traditional SEO.
The 37% figure is worth examining further. More than one in three consumers now report starting a search with an AI tool. Daily AI search users in the U.S. doubled from 14% to 29.2% of the adult population between early and late 2025. These are not marginal shifts. They represent a fundamental change in how a large portion of the consuming public approaches information retrieval.
What This Data Actually Means for Your Digital Strategy
SEO Is Not Dead — But It Cannot Be Your Only Strategy
The data from this study does not support the conclusion that traditional search optimization is obsolete. Google still processes billions of queries daily. Its advertising ecosystem is still the most sophisticated and high-volume digital marketing channel on the planet. For transactional and navigational queries, Google remains the dominant platform by a wide margin.
What the data does support is the conclusion that SEO, practiced in isolation, is an increasingly incomplete strategy. If 28% of global information-seeking queries are now being processed by AI assistants — and that percentage is growing — then a content strategy that optimizes exclusively for Google rankings is structurally unable to reach a substantial and growing portion of the audience you are trying to serve.
The framework that has emerged to address this is sometimes called Generative Engine Optimization (GEO) or, in slightly different formulations, Answer Engine Optimization (AEO) or LLM Optimization (LLMO). Whatever you call it, the underlying practice is the same: creating content and establishing authority signals in ways that cause AI systems to cite, recommend, or incorporate your brand, products, or expertise in their generated responses.
How AI Assistants Decide What to Surface
Understanding why GEO matters requires understanding how large language models construct answers. Unlike traditional search engines, which rank pages based on signals like backlinks, keyword relevance, and technical performance, LLMs generate answers by drawing on their training data and, for real-time models, live retrieval augmentation.
Several factors appear to influence whether a brand or piece of content is cited in AI-generated responses:
Entity recognition and repetition: Brands and sources that appear consistently and authoritatively across the web — in news coverage, industry publications, reviews, and original research — are more likely to be incorporated into AI-generated answers. This is not entirely different from traditional SEO authority-building, but it emphasizes breadth of citation across diverse sources.
Factual density and specificity: AI assistants favor content that contains specific, verifiable facts, data points, and original research. Generic content that restates common knowledge provides little value to an AI model already trained on that knowledge. Content with unique data — proprietary studies, original surveys, first-party analytics — is more likely to be cited.
Structured formatting: Clear headings, bullet-point summaries, definition-style explanations, and FAQ formats make it easier for AI systems to extract specific information for use in a synthesized answer. This structural preference aligns with older web accessibility best practices, which is useful for teams that have already invested in those areas.
Recency and freshness signals: Platforms like Perplexity that perform real-time retrieval place significant weight on content recency. For topics that are evolving quickly — like AI adoption statistics, market share data, or regulatory changes — regularly updated content has a meaningful advantage.
Brand authority across the broader web: AI models do not evaluate only the page where a specific piece of content lives. They draw on signals from across the web: Wikipedia entries, Wikidata records, press mentions, LinkedIn profiles, podcast appearances, industry awards, and user reviews. A brand with a thin web footprint is less likely to be cited regardless of how well its website content is technically structured.
The Traffic Quality Argument
One counter-narrative to the “AI is taking traffic” concern is worth taking seriously: the traffic that does arrive from AI platforms tends to convert significantly better than traditional organic search traffic. Analysis from Semrush indicates that referral visitors arriving from AI chatbots convert at approximately 4.4 times the rate of visitors from classic organic search results.
This conversion premium likely reflects intent: users who follow a link from an AI-generated answer have already received a synthesized overview of the topic and have self-selected to learn more. They are further along the decision journey than a user who just typed a broad keyword into Google. For e-commerce, lead generation, and subscription products, this quality differential is commercially significant.
A brand that loses 20% of its organic search traffic but captures a proportional share of higher-converting AI referral traffic may, in some scenarios, see flat or positive revenue outcomes even as raw traffic volume declines. This does not make the traffic decline irrelevant, but it complicates the simple narrative that “less search traffic equals less business.”
Practical Steps for Marketers Navigating This Transition
Audit your current AI visibility. Systematically test how your brand appears in responses from ChatGPT, Perplexity, Gemini, and Claude for your highest-priority queries. Note whether you are cited, whether the citations are accurate, and whether competitors are featured more prominently. This audit establishes a baseline for tracking GEO progress.
Build original, citable data assets. Commission original research, publish proprietary data, or analyze first-party data from your customer base. Data points that cannot be found elsewhere on the web are exactly what AI models need when they want to include a credible, specific claim in an answer. The Graphite.io study being discussed in this post is a perfect example: it produced a specific, citable figure (56%, 45 billion sessions) that appears across dozens of subsequent articles, analyses, and AI-generated responses.
Establish and optimize your brand’s entity presence. Ensure that your brand has a well-maintained Wikipedia page (if warranted by size), a complete and accurate Google Business Profile, detailed LinkedIn company and personal profiles for key leaders, consistent NAP (name, address, phone) information across directories, and active participation in the industry publications and review platforms that AI models are likely to draw on.
Implement structured data comprehensively. Schema markup for your organization, products, articles, FAQs, and people provides machine-readable signals that AI systems can use to understand and accurately represent your content. This is particularly important for FAQs, how-to content, and product specifications.
Separate your AI referral traffic in analytics. Set up UTM parameters and referral source filters in your analytics platform to distinguish traffic arriving from ChatGPT, Perplexity, Gemini, and Claude. Understanding the volume, behavior, and conversion rates of AI-referred traffic separately from other sources is a prerequisite for measuring the ROI of any GEO investment.
Update and maintain high-performing informational content. Pages that currently rank well for informational queries are candidates for AI citation precisely because they already have authority signals. Refreshing these pages with current data, adding FAQ sections, and improving their factual specificity can increase their likelihood of being surfaced in AI-generated answers.
Think in terms of mentions, not just rankings. Traditional SEO success is measured in rankings — position 1, 2, 3 in a results page. AI visibility is measured in mentions — does an AI system cite your brand, data, or content when answering questions relevant to your business? This is a meaningfully different measurement paradigm, and it requires different tracking tools and success metrics.
The Bigger Picture: Discovery Is Expanding, Not Shrinking
Perhaps the most important framing in the Graphite.io study is not the 56% figure itself, but the context in which it appears: total usage across search engines and AI assistants has grown 26% globally since 2023.
This is not a story of substitution. It is a story of expansion. People are seeking information, making decisions, and exploring ideas more than they ever have before. The total addressable market for information products — which is ultimately what both search engines and AI assistants compete in — is growing, not contracting.
For marketers and publishers, this should be motivating rather than threatening. The opportunity is not to defend a shrinking territory. It is to establish presence across an expanding landscape of discovery channels. Brands that approach the AI transition as a threat to be minimized will likely do worse than brands that approach it as a new channel to be optimized.
The analogy to social media’s emergence in the late 2000s and early 2010s is instructive. At that time, brands could either treat social media as an unwanted disruption to established advertising models, or invest early in building authentic presence and audiences on the new platforms. The brands that invested early — even imperfectly, even without clear ROI metrics — built advantages that were very difficult for late entrants to replicate. The AI search transition has a similar structure: early presence is easier to establish, and the compounding advantages of being a trusted, frequently-cited source in AI systems are likely to be significant.
The data tells a clear directional story: AI assistants are not replacing search engines, but they are becoming essential co-equal channels in how people find information, evaluate options, and make decisions. Ignoring either channel means willingly accepting invisibility with a measurable and growing portion of your potential audience.
Frequently Asked Questions
What does it mean that AI assistants equal 56% of global search engine volume?
The 56% figure comes from a study by Graphite.io CEO Ethan Smith, which measured both web traffic and mobile app usage across the five largest AI platforms (ChatGPT, Gemini, Perplexity, Grok, and Claude) and compared that total to sessions across the six largest traditional search engines. The result is that AI platforms generate approximately 45 billion monthly sessions globally, which is equivalent to 56% of total search engine session volume. This does not mean AI has 56% market share — traditional search still processes far more total queries. It means the volume of AI usage, when properly measured to include mobile apps, is proportionally much larger than most prior estimates suggested.
Why is the U.S. AI search figure lower (34%) than the global figure (56%)?
The U.S. figure of 34% reflects the fact that the United States has extremely high per-capita traditional search activity, which makes the AI-to-search ratio lower on a relative basis. However, in absolute growth terms, the U.S. is actually one of the fastest-growing AI markets — U.S. AI usage grew approximately 300% year-over-year by December 2025. The lower percentage is a function of a large denominator (very high U.S. search volume), not low AI adoption.
Why have previous studies underestimated AI usage by 4–5x?
Most prior comparisons between AI tools and search engines relied exclusively on web traffic measurement tools, which track visits to .com websites. However, approximately 83% of global AI usage and 75% of U.S. AI usage happens through native mobile apps, not web browsers. Web analytics tools cannot directly measure in-app activity. When you measure only website visits, you are seeing approximately 17–25% of total AI sessions. Multiplying the visible web traffic by 4–5x gives a more accurate picture of actual AI usage.
Which AI platform is most used globally?
ChatGPT dominates with approximately 89% of all global AI assistant sessions. It became the most downloaded app globally in 2025, surpassing TikTok. ChatGPT users also have significantly longer session durations (averaging 13+ minutes per session) compared to traditional search users on Google (approximately 6 minutes). The remaining 11% of AI sessions is divided among Gemini, Perplexity, Grok, and Claude, each serving somewhat different use cases and user segments.
Is Google’s dominance declining?
Google’s share of combined search engine plus AI assistant sessions fell from approximately 89% in 2023 to about 71% in Q4 2025, according to the Graphite.io study. However, within the traditional search engine category alone, Google still holds approximately 86–90% of global market share. Google’s position in transactional queries — which drive advertising revenue — remains near 90%. The decline is real, but it is primarily happening in informational queries rather than in the commercial queries that generate the majority of advertising value.
What is the difference between 56% and 28% in this study’s findings?
The 56% figure represents all AI assistant sessions as a percentage of search engine volume — this includes “asking” (information-seeking), “doing” (task completion), and “expressing” (creative generation) prompts. The 28% figure isolates only “asking” behavior — the subset of AI interactions that most directly resemble a traditional search query. OpenAI research estimates that approximately 52% of all ChatGPT prompts are information-seeking in nature. The 28% figure is generally more relevant for search marketers, as it measures the behavioral overlap between AI usage and traditional search.
What is GEO, and how is it different from SEO?
SEO (Search Engine Optimization) refers to the practice of structuring and promoting content to rank highly in traditional search engine results pages (SERPs), primarily Google. GEO (Generative Engine Optimization) refers to the practice of structuring content and building authority signals in ways that cause generative AI platforms — such as ChatGPT, Perplexity, Gemini, and Claude — to cite, reference, or incorporate your content in their generated answers. The key difference is the output format: SEO aims for a ranked position in a list of links, while GEO aims for citation within a synthesized narrative answer. The underlying requirements — authoritative content, factual accuracy, structured data, strong entity recognition — overlap significantly, but GEO places additional emphasis on original data, factual density, and breadth of web presence.
Should I stop investing in SEO and focus only on GEO?
No. Traditional search still processes the majority of information-seeking queries globally, and Google retains roughly 86–90% of search market share as of 2025. For transactional queries — those with commercial intent — Google’s dominance is even more pronounced. The evidence-based recommendation is to pursue both SEO and GEO as complementary strategies. Content that performs well in traditional search (authoritative, well-structured, factually rich) also tends to perform well in AI-generated answers. The marginal investment in GEO-specific optimizations — FAQ sections, structured data, original research, entity optimization — is relatively low for brands already investing in quality content.
Which AI platforms are most likely to send traffic to websites?
Perplexity has the highest outbound referral rate among major AI platforms — it is designed to visibly cite its sources and send users to those sources. Claude’s outbound clicks tend to go to search engines and research institutions, reflecting its user base’s deep-research workflows. ChatGPT’s referral behavior varies depending on whether the user is in a browsing-enabled session. Gemini, integrated into Google Search as AI Overviews, often surfaces information within the SERP itself rather than sending clicks through to third-party sites. For traffic generation specifically, Perplexity optimization is the highest priority after ChatGPT in terms of likely referral volume.
How does AI citation traffic convert compared to traditional search traffic?
Analysis from Semrush indicates that visitors arriving via AI chatbot referrals convert at approximately 4.4 times the rate of visitors arriving via traditional organic search. This premium likely reflects the fact that AI-referred users have already received a synthesized summary and are actively seeking deeper engagement — they are further along the decision journey than a user who entered a broad keyword. For e-commerce, lead generation, and subscription services, this conversion premium is commercially significant and can partially offset declines in raw organic traffic volume.
How should I measure AI search visibility for my brand?
There is no single universally adopted tool for measuring AI visibility as of early 2026, though several platforms are developing solutions. The most practical current approach involves: (1) manually auditing AI platform responses for your priority queries on a regular cadence; (2) setting up separate referral source tracking in your web analytics to isolate and monitor traffic from ChatGPT, Perplexity, Gemini, and Claude; (3) monitoring brand mentions across AI-adjacent web properties (publications, forums, and review sites that AI models tend to draw on); and (4) tracking whether your content is appearing in Google AI Overviews for target keywords using tools like Semrush’s AI Overviews tracker or similar features from SEO platforms.
What types of content perform best in AI-generated answers?
Content that performs best in AI citations tends to have several characteristics: original, specific data points that are not available elsewhere; clear, structured formatting with explicit headings and FAQ sections; factual density with verifiable claims and citations; expert authorship with demonstrable credentials; recency for time-sensitive topics; and broad web authority established through mentions across multiple independent sources. Long-form, comprehensive content that covers a topic thoroughly tends to outperform thin content in AI citations, as AI models favor sources that demonstrate depth of coverage.
Is AI search usage still growing?
Growth patterns differ by geography. Globally, AI usage growth appeared to plateau around July 2025. In the United States, however, growth remained strong, with usage up approximately 300% year-over-year by December 2025. Daily AI search users in the U.S. roughly doubled from 14% to 29.2% of the adult population during 2025. The expectation among most analysts is that international markets will follow a similar adoption curve, with growth resuming as AI capabilities improve and awareness increases in markets where adoption currently lags the U.S.
What does this mean for content publishers and media sites?
For content publishers that rely on informational content to attract organic search traffic, the trends outlined in this study represent a material challenge. AI Overviews on Google have been shown in multiple studies to reduce organic click-through rates — in some cases by 40–60% for affected informational queries. At the same time, AI platforms like Perplexity that cite sources can become new referral channels for publishers cited in their answers. The practical recommendation for publishers is to (1) invest in original reporting and data that AI systems cannot reproduce from common knowledge, (2) ensure deep structured-data implementation that facilitates accurate AI citation, and (3) monitor and attempt to establish presence in Perplexity’s citation ecosystem, which currently sends proportionally more traffic to publishers than ChatGPT.
How is Google responding to the rise of AI assistants?
Google has not been passive in response to the rise of AI assistants. Its primary strategic responses include: integrating Gemini into Google Search as AI Overviews; deploying “AI Mode” in Google Search, which provides a full conversational AI interface within the Google ecosystem; expanding Gemini’s integration across Android devices, Google Workspace, and third-party applications; and investing heavily in Gemini’s capabilities to match or exceed ChatGPT’s quality on information-seeking tasks. The effect of these moves is that some portion of the usage that would previously have migrated fully to standalone AI apps stays within the Google ecosystem — making it harder to measure as “AI vs. Google” and more accurately framed as “traditional Google vs. AI-augmented Google.”
Where This Leads: An Honest Assessment
The Graphite.io study does one thing very well: it establishes that the AI assistant space is much larger, by every meaningful measure, than the industry had previously documented. Forty-five billion monthly sessions is not a rounding error. The fact that those sessions were largely invisible to standard web analytics tools is a data infrastructure problem, not evidence that the activity was not happening.
The 56% equivalence figure is a useful attention-getter, but the strategically important numbers are these: 28% of global information-seeking behavior is now happening on AI platforms; Google’s share of combined discovery activity has fallen from 89% to 71% in two years; U.S. AI usage grew 300% in a single year; and 37% of U.S. consumers now start information searches with AI, not Google. Together, these figures describe a landscape that has already changed substantially — and is likely to keep changing.
The appropriate response is not to abandon what is working. Traditional search optimization built on quality, authority, and well-structured content continues to produce results, and the skills developed for SEO transfer meaningfully to GEO. The appropriate response is to extend your strategy to the new channels where your audience is increasingly spending their information-seeking time, and to build the kind of original, authoritative, factually specific content that performs across both environments.
The brands and publishers that do this work now, before AI optimization becomes standard practice, will establish compounding advantages in AI visibility that will be difficult for later entrants to close. The window is open. The data — not projection, not forecast, but measured usage data from 2025 — shows why it matters.
Source: Ethan Smith, Graphite.io; “AI Assistants Now Equal 56% of Global Search Engine Volume: Study,” Search Engine Land.
About ALM Corp
ALM Corp is a digital strategy and marketing firm that helps businesses build measurable visibility across both traditional search engines and emerging AI discovery platforms. As the landscape of online search continues to evolve — with AI assistants now generating sessions equivalent to 56% of global search engine volume — the gap between brands that understand the new rules of digital visibility and those that do not is widening rapidly.
ALM Corp’s integrated approach combines technical SEO, content strategy, structured data implementation, and Generative Engine Optimization (GEO) to ensure that client brands appear not only in Google’s top results but also in the AI-generated answers that an increasingly large share of their target audience is relying on. Whether your priority is ranking in Google AI Overviews, being cited by ChatGPT and Perplexity, or building the kind of authoritative content that AI models trust as a reliable source, ALM Corp delivers strategies grounded in current data — not speculation.
In a world where search is fragmenting across multiple discovery channels, ALM Corp ensures your brand is present and influential across all of them.
Contact ALM Corp to discuss how your current digital strategy maps to this evolving landscape.
