AI Is Already 56% the Size of Global Search

AI Has Quietly Become Half the Size of Global Search. And Most of the Data Was Being Ignored.

A sweeping new analysis has upended how analysts and marketers have been measuring the rise of artificial intelligence as an information platform. The report, which combines web visits with mobile app sessions for the first time in a large-scale comparison, finds that AI is already 56% the size of global search by monthly sessions, a figure that is four to five times larger than previous estimates that relied solely on web traffic data.

The findings reframe one of the most hotly debated questions in technology: is AI killing search? The short answer, according to the data, is no. But it is transforming the total landscape of how people find and interact with information, at a pace that has surprised even the researchers behind the study.

Why Previous Estimates Were So Wrong

Most prior comparisons of AI versus search traffic made a fundamental methodological error: they only counted web visits. That approach missed the single largest channel through which people now access AI. According to the analysis, 83% of AI usage globally occurs on mobile apps, with only 17% taking place on the web. In the United States, the mobile share is 75%.

By ignoring app sessions, earlier projections were undercounting total AI usage by a factor of four to five. This explains why claims about AI's rapid ascent were simultaneously dismissed as hype and, in fact, understated in absolute terms.

"We originally sought to evaluate our hypothesis that most claims about the usage of AI were overstated, and that usage of AI was smaller and growing more slowly. However, upon closer inspection of the data, we found that not only are the claims that AI may soon overtake search not overstated, they are understated." â Research team, as published in the study

A second flaw in prior analyses was scope. Most comparisons pitted ChatGPT's website against Google Search alone, ignoring the broader landscape of competing large language models such as Gemini, Perplexity, Grok, and Claude. Those additional platforms now account for 11% of global AI usage and 14% of US AI usage.

The data source underpinning the study is Similarweb, validated against first-party signals. The researchers calculated a median Pearson correlation coefficient of 0.86 between Similarweb monthly session data and first-party data from Google Search Console and Google Analytics, considered a very high correlation.

Search Is Not Dying. The Pie Is Growing.

One of the most important conclusions of the study is one that runs counter to the dominant media narrative: search usage has not declined. Google continues to handle over five trillion searches annually, a figure the company itself confirmed publicly in March 2025. When AI Asking sessions (defined as prompts seeking information or advice) are combined with traditional search engine visits, total information-seeking activity worldwide has increased by 26% comparing 2025 with 2024.

The study draws a pointed parallel with the mobile app revolution of the early 2010s. When Apple introduced the App Store in 2008, technology commentators predicted the death of the web. Wired magazine declared in September 2010 that "The web is dead. Long live the internet." The web did not die. It grew, and mobile usage grew on top of it. The researchers argue the same dynamic is now playing out with AI and search.

"We already see more than 5 trillion searches on Google annually." â Google, public announcement, March 2025

This phenomenon, which the report labels zero-sum bias, describes the cognitive tendency to assume that gains in one technology must come at the expense of another. The emergence of Netflix destroyed Blockbuster. The internet killed the Yellow Pages. But the growth of mobile did not kill the web, and the data suggest AI is not killing search either.

ChatGPT Commands 20% of Global Search Traffic

When search-related AI usage (Asking prompts only) is factored in, ChatGPT now accounts for 20% of search-related traffic worldwide and 12% in the United States. Google's global market share in search and discovery has declined from 89% in 2023 to 71% in Q4 2025. In the US, that figure has moved from 88% to 75%.

ChatGPT remains the dominant AI platform by a wide margin. It holds 89% of the global AI market share and 86% in the United States. The next closest competitors, Gemini, Perplexity, Grok, and Claude, collectively account for the remaining share.

The study draws on session data rather than unique visitors or monthly active users, a deliberate choice. Because many users access multiple LLM products across both web and mobile, unique visitor counts would involve unresolvable duplication. Sessions represent actual usage behaviour more cleanly.

This shift in how AI sessions are being discussed also has implications for how we think about software development itself. Tools that once required specialist engineering teams to build are now being assembled through conversational prompts. As we explored in our coverage of how vibe coding is reshaping the way software gets built, the line between AI as a tool and AI as an operator is blurring fast.

The Anatomy of an AI Prompt

Not all AI usage is search-equivalent. The study draws on a Harvard and OpenAI analysis of over one million deidentified ChatGPT prompts, categorising them by user intent. The breakdown is instructive.

• Asking (51.6%): Seeking information or advice. Examples: "What was the inflation rate last year?" or "What should I look for when choosing a health plan?" These are the prompts most comparable to a search query.
• Doing (34.6%): Requesting the model to perform a task. Examples: "Rewrite this email", "Draft a report", "Write a Dockerfile". These prompts have no equivalent in traditional search.
• Expressing (13.8%): Conversational or expressive messages that are neither information requests nor task instructions.

When only Asking prompts are counted in the AI-versus-search comparison, the AI side of the ledger shrinks meaningfully. Search-related AI usage represents 28% of global search volume worldwide and 17% in the United States. This is the appropriate apples-to-apples comparison, though the researchers note it is still an upper bound, given that many Asking prompts address questions that users would never have typed into a search engine at all.

The Asia-Pacific Picture

Perhaps the most striking finding in the entire dataset is one that receives the least attention in Western technology coverage: AI usage outside the United States is seven times larger than within it. Global AI sessions reached 45 billion per month, compared to 5.4 billion in the US alone. That arithmetic places the rest of the world at approximately 39.6 billion monthly sessions, a figure that dwarfs domestic American adoption.

Asia-Pacific is a central driver of this gap. The region has some of the world's highest mobile-first internet populations, making it structurally primed for AI adoption via apps rather than web browsers. Markets such as India, Indonesia, and the Philippines have large, young, mobile-native user bases that access AI tools through smartphones as their primary computing device, precisely the modality that previous measurements were ignoring.

China presents a distinct picture. Baidu and Yandex are included in the worldwide search engine dataset but excluded from US figures. China's domestic AI ecosystem, anchored by models such as Baidu's ERNIE, Alibaba's Qwen, and ByteDance's Doubao, operates largely outside the ChatGPT-centric data in this study. The 89% global market share attributed to ChatGPT almost certainly overstates its position within mainland China, where access to the platform is restricted and domestic alternatives dominate.

For policymakers and businesses in the region, the implications are significant. Vietnam has already enacted Southeast Asia's first AI law, signalling that regulators are beginning to formalise oversight of the very platforms driving these usage numbers. Meanwhile, China has placed AI at the centre of its next five-year plan, a structural commitment that will only accelerate domestic AI session volumes in data sets that capture Chinese platforms.

The commercial opportunity implicit in these numbers is considerable. AI-driven discovery is now a genuine channel for reaching consumers across Asia, and as we noted in our analysis of how AI has already changed how Asia shops, the shift from search-driven commerce to AI-assisted purchase decisions is further advanced in this region than most Western brands have yet appreciated.

AI Usage: US vs Worldwide at a Glance

What This Means for Search, Marketing, and Strategy

For digital marketers and strategists, the study dismantles several comfortable assumptions at once. The idea that AI growth is primarily a desktop, tech-savvy behaviour is wrong: 83% of global AI usage is on mobile apps. The idea that Google is losing traffic is wrong: its absolute visit volumes are stable. The idea that AI is a niche complement to search is wrong: it is now a substantial and growing share of total information-seeking behaviour.

What is shifting is market share within a growing total. Google held 89% of the search and discovery market in 2023. By Q4 2025, that had fallen to 71% globally, not because fewer people are using Google, but because a larger and larger cohort of information-seekers is also using ChatGPT, Gemini, Perplexity, and their competitors.

The plateau in worldwide AI growth since July 2025 is worth watching. The study notes that global sessions have levelled off, even as US usage continues to climb sharply, up 300% year-on-year in December 2025. Whether that plateau represents a ceiling or a pause ahead of further growth driven by new model releases and device integrations remains an open question.

Frequently Asked Questions

How big is AI compared to Google Search right now?

As of late 2025, total AI sessions worldwide (across ChatGPT, Gemini, Perplexity, Grok, and Claude, combining web and mobile app usage) are equivalent to 56% of global search engine sessions. In the United States, that figure is 34%. When only search-equivalent "Asking" prompts are counted, AI represents 28% of global search volume and 17% in the US.

Is Google Search traffic declining because of AI?

No. Google confirmed in March 2025 that it handles over five trillion searches annually, consistent with Similarweb data. Search usage has been flat to growing over the past six years. What has changed is market share within an expanding total: Google's share of all information-seeking activity has declined from 89% in 2023 to 71% in Q4 2025 as AI platforms have taken a growing slice of a larger overall pie.

Why is AI usage so much larger outside the United States?

Global AI sessions are roughly seven times larger than US sessions, totalling 45 billion monthly versus 5.4 billion in the US. This reflects the mobile-first nature of AI adoption, particularly across Asia-Pacific, where large populations access ChatGPT and other LLMs through smartphone apps. Previous estimates that only counted web traffic systematically undercounted these markets.

YOUR TAKE

We cover the story. You tell us what it means on the ground.