Pew, McKinsey, G2, Adobe and Comscore measure different things. Together they show AI is mainstream, but not its universal share of search.
<aside role="note" class="my-8 border-y border-[#D9C7A7] bg-[#FFF8EA] px-5 py-5 text-[#171C21]"><strong>Historic record:</strong> This August 29 analysis has been superseded by <a href="/articles/ai-search-reaches-billions-no-single-market-share" class="font-semibold text-[#2448B8] underline underline-offset-2">AI search reaches billions. That still does not tell us its market share.</a> The original analysis remains available below for transparency.</aside>
Around two in five US adults say they use AI chatbots to search for information. Half of US consumers in another study said they intentionally seek AI-powered search. Seven in ten surveyed B2B software buyers said they use AI chatbots somewhere in their research process. McKinsey has also estimated that AI summaries appear on about half of Google searches.[1] [2] [4]
Those numbers sound as if they should answer a simple question: how big is AI search?
They do not, because they are not measuring the same thing. One measures people, another uses a broader definition that includes AI inside Google, a third covers a specific B2B buying population, and another estimates the incidence of an AI feature within conventional search. Combining them would create false precision.
The more defensible conclusion is also more useful: <strong>AI has not replaced search. AI is becoming part of search, and its commercial importance is not captured by one market-share percentage.</strong>
<h2>One market, at least three denominators</h2>
The size of AI search depends first on the unit being counted. Three measurement frames answer three different questions.
<strong>Person-level adoption</strong> asks what percentage of people use AI to find information. <strong>Occasion-level migration</strong> asks what percentage of information-seeking activity has moved from traditional search to standalone assistants. <strong>AI-mediated exposure and decision use</strong> asks how often people encounter an AI-generated answer or use AI during research, comparison and selection.
These frames can all rise at the same time without moving at the same rate. A person who performs 50 Google searches and five ChatGPT searches in a week is an AI-search user, but only a small share of that person’s combined activity occurred in ChatGPT. If some of the Google searches also triggered AI Overviews, the person was exposed to more AI-mediated search without leaving Google.
This is the same denominator problem that complicates AI visibility measurement. As the AEO Updates analysis <a href="/articles/ai-visibility-no-common-denominator">AI visibility has no common denominator</a> argues, a percentage is meaningful only when its population, prompt universe, system and unit of analysis are visible.
<h2>Two surveys bracket adoption, not search share</h2>
Pew Research Center surveyed 5,119 US adults from 17 to 23 February 2026. It found that 49% had ever used an AI chatbot and 42% said they used chatbots to search for information. Searching for information was the most common use Pew measured.[1]
McKinsey used a broader frame. Its AI Discovery Survey covered a representative panel of 1,927 US consumers in August 2025 and included both standalone applications such as ChatGPT, Gemini, Copilot, Perplexity and Claude and Google AI Overviews. Half said they intentionally sought AI-powered search.[2]
These results support a useful description, not a pooled estimate: <strong>depending on the definition, roughly two in five to one in two US adults or consumers report using AI for information search in some form.</strong> Pew’s 42% and McKinsey’s 50% should not be averaged because their populations, wording and product boundaries differ.
They also do not show that two in five or one in two search occasions have moved away from Google. They measure whether people use AI search, not how many times they use it or which other search tools they continue to use.
<h2>Behaviour shows displacement without establishing market share</h2>
A July 2026 preprint by Qiaoni Shi, Kai Zhu and Kai Gu adds a different form of evidence. The researchers analysed URL-level Comscore US desktop clickstream data from October 2024 to July 2025. Their design used three expansions of ChatGPT Search access and compared 3,882 treated households with reweighted controls.[3]
The study estimated that broader ChatGPT Search access reduced weekly Google, Bing and Yahoo query loads by 9.4% relative to the pre-expansion mean among treated households. The decline was larger after sustained exposure and concentrated in informational categories.[3]
That is meaningful evidence of substitution. It is not evidence that ChatGPT owns 9.4% of search. The figure is a treatment effect within a historical US desktop panel, not a census of every search occasion across devices, platforms and populations. The paper is also a preprint, and its panel-wide identification of search-enabled ChatGPT sessions relies on an audited but imperfect endpoint proxy.
The distinction matters because adoption, displacement and market share are not interchangeable. The available research does not yet provide a comprehensive census that can support a universal estimate for standalone AI’s share of intentional digital information search. A proposed 10% or 15% number would be a planning hypothesis, not a measured fact.
<h2>Commercial research is moving faster than the average</h2>
The strongest marketer signal is not total search volume. It is the concentration of AI use in research and decision contexts.
McKinsey reported that approximately 40% to 55% of consumers in seven named sectors, including consumer electronics, travel, wellness, apparel, beauty and financial services, were using AI-based search to make purchasing decisions. Among AI-powered-search users, 44% described it as their primary and preferred source of insight, compared with 31% for traditional search. The denominator for that preference result is AI-powered-search users, not all consumers.[2]
G2’s March 2026 online survey covered 1,076 B2B decision-makers involved in software purchases across North America, EMEA and APAC. It found that 71% relied on AI chatbots somewhere in software research and 51% started with an AI chatbot more often than Google. Yet 80% still used Google somewhere in the buying journey, and 61% said they used AI search alongside Google.[4]
Adobe’s US shopping studies point in the same direction within a different category. In March 2025, 39% of 5,000 surveyed consumers said they had used generative AI for online shopping. Among the survey’s AI-shopping users, research, product recommendations and deal finding were leading tasks. A later 5,000-consumer wave reported similar adoption at 38%.[5] [6]
None of these studies measures the general population in the same way. McKinsey covers named commercial categories, G2 covers global B2B software buyers, and Adobe covers US online shopping. They should not be combined into one number. Together, however, they support a bounded interpretation: <strong>AI use appears particularly consequential where a question requires synthesis, comparison, constraints, trade-offs or a recommendation.</strong>
AI does not need to win most searches to become commercially important. It needs to mediate the decisions that matter.
<h2>Google is transforming search from within</h2>
The boundary becomes even harder to draw because AI search is not limited to standalone assistants.
Pew found that 60% of US adults said they had read AI summaries at the top of search-engine results. That is person-level reported exposure, not the percentage of searches that triggered a summary.[1]
McKinsey separately estimated through trend analysis that about half of Google searches already had AI summaries. The methodology behind that estimate is not disclosed on the article page, so it should be treated as McKinsey’s estimate rather than an independently reproducible market measure.[2]
Google’s own documentation describes AI Overviews as snapshots that summarise key information with links and AI Mode as a search experience for nuanced questions, reasoning, complex comparisons and follow-ups. Google also says AI features are part of Search, while existing SEO fundamentals remain relevant and no special technical requirements guarantee inclusion.[7]
This creates two simultaneous changes. Some information-seeking is migrating from traditional search engines to standalone assistants. At the same time, traditional search is becoming more answer-led. The first changes where the session begins. The second changes what happens before a person decides whether to click.
<h2>Referral traffic captures the handoff, not the whole influence</h2>
The Comscore study found that 5.2% of 409,133 observed ChatGPT conversation sessions produced at least one clean outbound referral, compared with 31.1% of Google queries. This is not a click-through rate per displayed link. Its denominator is the complete information-seeking occasion reconstructed by the researchers.[3]
An absorbed session could mean the answer satisfied the user, the task was abandoned, a visit lost its referrer or something else occurred. The paper does not observe task completion or consumer welfare. It does show why direct referral traffic captures only one part of the AI-search journey.
Adobe’s behavioural retail data provides the complementary signal. Generative-AI referrals to US retail sites rose 4,700% year over year in July 2025, but Adobe still described the channel as modest compared with paid search or email. Fast growth from a small base does not establish large channel share.[6]
The two findings can coexist. AI may influence discovery, comparison and shortlisting without creating a directly attributable click, while the smaller referral stream grows quickly. The AEO Updates analysis <a href="/articles/ai-search-website-downstream-brainlabs">AI search may be moving the website down the funnel</a> examines the related possibility that a website increasingly receives the handoff for verification, pricing, service or transaction rather than every earlier research visit.
<h2>There is no single answer to “How big?”</h2>
The evidence supports a measurement scorecard rather than one headline market-share figure.
<div class="overflow-x-auto"><table><thead><tr><th>Question</th><th>Best current evidence</th><th>Boundary</th></tr></thead><tbody><tr><td>How many US adults use chatbots for information search?</td><td>42% in Pew’s February 2026 survey</td><td>Self-reported person-level use, not occasion share</td></tr><tr><td>How many US consumers intentionally seek broadly defined AI-powered search?</td><td>50% in McKinsey’s August 2025 survey</td><td>Includes standalone assistants and Google AI Overviews</td></tr><tr><td>Does wider ChatGPT Search access reduce traditional search?</td><td>9.4% fewer weekly queries in the Comscore study’s treated desktop households</td><td>Treatment effect, not universal market share</td></tr><tr><td>How common is AI use in B2B software research?</td><td>71% in G2’s March 2026 global buyer survey</td><td>Category-specific professional population</td></tr><tr><td>How common is generative AI in US online shopping?</td><td>38% to 39% across two Adobe survey waves</td><td>Shopping context; separate survey waves</td></tr><tr><td>How many US adults report reading AI search summaries?</td><td>60% in Pew’s survey</td><td>Ever-exposed people, not query incidence</td></tr><tr><td>How often did observed ChatGPT sessions create an outbound referral?</td><td>5.2% in the Comscore desktop panel</td><td>Conversation-session ratio, not link CTR or task completion</td></tr></tbody></table></div>
The rows are deliberately incomparable. Each answers a different management question.
<h2>AEO Updates Takeaway</h2>
Marketers should stop asking for one universal AI-search percentage before deciding whether the channel matters. A more useful measurement declaration separates person-level adoption, occasion-level migration, exposure to AI-generated answers, decision-stage use, observable referrals and downstream business outcomes.
For AEO programmes, the practical question is not simply whether AI owns 10%, 20% or 50% of “search”. It is which customer questions are now mediated by AI, whether the brand appears and is accurately represented, whether it enters the consideration set and receives the preferred recommendation, what observable sources support those answers, and where a handoff produces qualified traffic, leads, purchases or revenue.
That requires a stable prompt and audience frame, separate platform reporting, visible denominators and repeated measurement. The current <a href="/state-of-aeo-2026">State of AEO 2026</a> provides the wider measurement framework.
The best available evidence does not support the claim that half of search has moved from Google to ChatGPT. It supports a more important conclusion: AI-assisted information-seeking is mainstream, AI is increasingly embedded inside search, and category-specific research suggests it is already influential in many high-consideration decisions.
<h3>References</h3>
[1] <a href="https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/" target="_blank" rel="noopener noreferrer">Pew Research Center, “Americans and AI 2026: Chatbots, Smart Devices and Views on Impact”</a>, 17 June 2026.
[2] <a href="https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/new-front-door-to-the-internet-winning-in-the-age-of-ai-search" target="_blank" rel="noopener noreferrer">McKinsey & Company, “New front door to the internet: Winning in the age of AI search”</a>, 16 October 2025.
[3] <a href="https://arxiv.org/abs/2607.07652" target="_blank" rel="noopener noreferrer">Qiaoni Shi, Kai Zhu and Kai Gu, “Answering Without Referring: How AI Search Rewrites the Web’s Economic Bargain”</a>, arXiv preprint, 8 July 2026.
[4] <a href="https://learn.g2.com/g2-2026-ai-search-insight-report" target="_blank" rel="noopener noreferrer">G2, “The Answer Economy: How AI Search Is Rewiring B2B Software Buying”</a>, April 2026.
[5] <a href="https://blog.adobe.com/en/publish/2025/03/17/adobe-analytics-traffic-to-us-retail-websites-from-generative-ai-sources-jumps-1200-percent" target="_blank" rel="noopener noreferrer">Adobe, “Traffic to U.S. retail websites from Generative AI sources jumps 1,200 percent”</a>, 17 March 2025.
[6] <a href="https://business.adobe.com/blog/generative-ai-powered-shopping-rises-with-traffic-to-retail-sites" target="_blank" rel="noopener noreferrer">Adobe, “Generative AI-powered shopping rises with traffic to U.S. retail sites up 4,700%”</a>, 21 August 2025.
[7] <a href="https://developers.google.com/search/docs/appearance/ai-features" target="_blank" rel="noopener noreferrer">Google Search Central, “AI features and your website”</a>, updated 10 December 2025.