The Global Divergence

The click-cost is now measured, not assumed

Pew Research Center ran the cleanest test available on what an AI summary actually does to a click, tracking 68,879 unique Google searches across a panel of 900 U.S. adults in March and April 2025. When no AI summary appeared, 15% of visits included a click on a traditional search result. When an AI summary appeared, that fell to 8%, and only 1% of visits included a click on a link inside the summary itself. More telling still, 26% of visits to a page with an AI summary ended the search session right there, against 16% for pages without one.

That is not a projection or an industry estimate. It is a measured behavior gap from a controlled panel study, and it is the strongest single data point in this study for a simple reason: the summary does not just steal the click, it often ends the search entirely. For a business whose visibility has always depended on being the blue link someone clicked, that is the finding to sit with first.

When an AI summary appeared, only 8% of visits included a click on a traditional search result, down from 15% when it didn't, and just 1% clicked a link inside the summary itself.

A widening reliance divide: flat West, accelerating Asia and Southern Europe

The Reuters Institute's Digital News Report 2026 gives the clearest cross-national read on where AI-answer habits are actually moving, and the picture is not one of steady global convergence. Weekly use of AI chatbots for news rose from 7% in 2025 to 10% in 2026 worldwide, reaching 16% among people under 35. But that global average hides a split: the United States, the United Kingdom, France, and Germany reported no year-on-year increase at all, while South Korea, Greece, and Spain each saw usage roughly double.

That divergence matters more than the global average does. A business built around US or Western European buyer habits is operating in a market where AI-chatbot reliance for information has, on this evidence, plateaued for now. A business with customers in South Korea, Greece, or Spain is operating in a market where that same habit is compounding fast. Treating either group the same way, on the strength of one worldwide percentage, would misread both.

The US, UK, France, and Germany reported no year-on-year increase in weekly AI-chatbot-for-news use, while South Korea, Greece, and Spain each roughly doubled theirs.

Trust hasn't caught up to habit

Growth in AI-chatbot use has not brought growth in trust to match it. Among the people who do use AI chatbots for news, the Reuters Institute found that 42% globally say they always or often click through to the original source, but that figure swings from 26% in Denmark to 56% in South Korea, a wide enough range that no single number describes the behavior well. Overall trust in AI chatbot answers sits at just 20% globally, and only 6% in the UK.

For a small business, this cuts two ways at once. Low trust means people who do encounter a synthesized answer are, in meaningful numbers, still checking the source behind it, which keeps being the cited, verifiable source valuable even inside the AI-answer layer. But low trust combined with rising use also means more of the buying journey is happening inside an answer a business cannot see, cannot audit, and was not asked permission to be part of.

Search still anchors discovery, and the AI census stays fuzzy

It is worth remembering how much of global discovery still runs through plain search. DataReportal's Digital 2026 Global Overview Report counts 6.04 billion people online worldwide, 73.2% of the global population, with 2.21 billion people still offline. Nearly one in three internet users aged 16 and older say they discover new products or services through search engines like Google and Bing, which remains one of the largest single discovery channels measured anywhere in the report.

The same report is honest about what it cannot count. It offers no verified combined total for the major chatbot platforms, and instead states a hedged estimate of more than one billion people using large language model or generative AI platforms monthly, built from extrapolation rather than an official combined figure (with ChatGPT alone estimated near one billion monthly active users). That gap between a hard search-discovery number and a hedged AI-usage estimate is itself a finding: the tool used to discover products at global scale is still, provably, search, while the AI-answer layer's true reach remains an informed guess even from one of the field's most cited data aggregators.

The fast-growth markets: real momentum, uneven proof

The fastest AI-reliance growth in this evidence set sits outside the US and Western Europe, but the proof behind it is thinner. India's internet base is closing in on a billion users: the Internet and Mobile Association of India and Kantar's ICUBE 2025 survey, covering more than 87,000 respondents across over 400 towns and 1,000 villages, counted 958 million active internet users, growing roughly 8% year over year, with the 548 million rural active users growing faster than urban ones. Forty-four percent of users reported having used AI-enabled features, though that figure bundles voice search, image search, chatbots, and AI filters into one composite number rather than isolating answer-engine use specifically. Short-video consumption reached 588 million people, or 61% of users, in 2025.

Southeast Asia tells a similar growth story through a different lens. Google, Temasek, and Bain's e-Conomy SEA 2025 survey of 7,200 consumers across the ten ASEAN markets found regional interest in general AI running at three times the global average, with 75% saying AI chatbots, translators, or image search helped them discover content and 30% saying AI helped them discover a brand or product they would otherwise have missed. Three in five still want a human confirmation step before acting. The region's digital economy is projected to surpass $300 billion in gross merchandise value in 2025, up 15% year over year, and among early enterprise adopters, 60% report growth from generative AI and 60% report more than 3x return from agentic AI.

Latin America's numbers rest on firmer ground. EMARKETER, an independent analyst firm rather than a report sponsor with a stake in the outcome, projects LATAM digital ad spend to reach $50.1 billion in 2026 and grow at a 12.3% compound rate to roughly $70.9 billion by 2029, with digital now 56.4% of total media spend, up from under 20% in 2015. Regional ecommerce is projected to exceed $215 billion in 2026, growing about 1.5 times the global average, making Latin America the world's second-fastest-growing region for digital ad spend next year.

All three regions are growing faster than the US and Western Europe on the measures available. But the India and Southeast Asia figures come from single-sponsor industry surveys, however well-resourced, while the Latin America figures come from independent market analysis. That distinction should shape how much weight a business puts on each one.

The SME gap: rising adoption, persistent distance from large firms

US small businesses have moved quickly on AI in a short window. The U.S. Chamber of Commerce's Technology Engagement Center, working with Teneo Research, found that almost 60% of US small businesses now report using AI for business operations, up from 23% in 2023, and 58% specifically use generative AI, up from 40% in 2024. The Chamber has an advocacy interest in a favorable adoption story, so this figure is best read as directional rather than independently audited, but the direction itself is not in question.

What that adoption curve does not close is the gap with larger firms. The OECD reports that the SME-versus-large-firm AI adoption gap runs more than three times wider than the gap seen for older digital technologies. Among the SMEs that do adopt generative AI, the payoff is real: 91% report efficiency gains, 76% report increased innovation, and over 60% report reduced staffing needs or access to new revenue streams. The lesson is not that small businesses are failing to adopt AI. It is that they are adopting it from further behind, on a technology where the visibility stakes of being behind are only now becoming clear.

The OECD finds the SME-versus-large-firm AI adoption gap runs more than three times wider than the gap seen for older digital technologies.

A nascent optimization discipline, and vendor data to read carefully

One academic paper has attempted to measure whether content can be deliberately optimized to win more citations inside generative-AI answers. The Generative Engine Optimization paper, accepted at KDD 2024, reports visibility gains of up to 40% in generative engine responses from specific optimization strategies, with the effect varying by domain. That is a genuine, peer-reviewed finding, and it establishes that answer-engine visibility is a measurable, influenceable thing rather than pure guesswork. But it is a single paper with no independent replication found, so the real-world size of that effect across industries and engines remains thinly evidenced.

Meanwhile, the numbers most often quoted about how often AI Overviews appear come from SEO vendor crawlers, not from Google itself. A widely cited compilation attributes to Ahrefs a figure that AI Overviews now appear on almost 55% of Google searches and reduce click-through by close to 35% across a sample of more than 300,000 keywords, alongside a SparkToro estimate that 58% of Google searches end with zero clicks at all. These figures were directly checked against their source and the attributions hold up. But they remain third-party crawler estimates, not data published by Google, a regulator, or a peer-reviewed source, which is why this study tiers them as contested rather than established. Use them as a directional signal about scale, not as a precise number to build a forecast on.

What the divergence means for where a business's attention terrain sits

Put together, the evidence supports one clear conclusion and rules out another. It rules out the idea of one global AI-search story that every business can plan against the same way. Reliance is flat in the US, the UK, France, and Germany, and doubling in South Korea, Greece, and Spain. Trust ranges from 6% in the UK to double digits elsewhere. Growth in India, Southeast Asia, and Latin America is real but measured with very different rigor from market to market.

What it supports is the case for treating attention as terrain rather than a single number: a map of where a specific market's search, social, short-video, and AI-answer behavior actually sits today, read against your own customers rather than a worldwide average built for headlines. No government body, platform, or academic source has yet published an independent census of how much of the buying journey the AI-answer layer actually controls in any one country. Until one exists, the defensible move for a small or mid-size business is to measure its own terrain directly rather than assume the global figure describes it.