AI Search Consumer Behavior in 2026: How People Research, Compare, and Choose

The market in six numbers

01

AI has reached mass-market distribution

OpenAI reported more than 900 million weekly active ChatGPT users in February 2026. In the United States, Pew Research Center found that 49% of adults use an AI chatbot and 44% specifically use ChatGPT. ChatGPT use has more than doubled from the 18% Pew measured in 2023.

Frequency matters as much as reach. Pew found that 24% of U.S. adults use AI chatbots daily: 8% about once a day, 12% several times a day, and 4% almost constantly. Another quarter use them several times a week or less. This is habitual behavior for a substantial group, not an occasional experiment.

These numbers establish distribution, not search-market share. They do not tell us what percentage of all searches now happen in ChatGPT, and they should not be compared directly with Google query volume. They show that conversational interfaces already have an audience large and frequent enough to influence discovery behavior.

02

People increasingly use ChatGPT as an adviser

OpenAI and its research collaborators classified 1.5 million consumer ChatGPT conversations without researchers reading the messages. Practical Guidance, Seeking Information, and Writing accounted for approximately 77% of conversations.

The mix changed materially over the study period. Seeking Information grew from 14% of messages in July 2024 to 24% a year later. Practical Guidance remained close to 29%, while Writing declined from 36% to 24%. Seeking Information included questions about people, current events, products, and recipes; Practical Guidance included personalized how-to advice, teaching, ideas, and recommendations.

The intent classification points in the same direction. Forty-nine percent of messages were categorized as Asking: seeking information or advice that helps the user become better informed or make a decision. Forty percent were Doing and 11% were Expressing. The model is still used to produce work, but its role as an adviser and decision-support interface is already central.

03

What consumers ask AI to do before they buy

Adobe surveyed 5,000 U.S. consumers in early 2025 and found that 39% had used generative AI for online shopping. Among those consumers, 55% used it to conduct research, 47% to receive product recommendations, and 43% to find deals.

The remaining uses reveal the shape of conversational demand: 35% used AI for gift ideas, 35% to find unique products, and 33% to create shopping lists. These are not simple navigational searches for a known website. They are questions about suitability, alternatives, constraints, price, and choice.

The practical unit of AI discovery is therefore often a buying situation rather than a keyword. A customer can combine category, budget, location, experience level, compatibility, urgency, and personal preference in one request. The answer can perform research, comparison, and shortlisting at the same time.

04

AI builds the shortlist more readily than it closes the sale

Current consumer evidence does not support the idea that most people want AI to make purchases autonomously. Gartner found that willingness to let AI make a purchase decision topped out at 11% across the lower-stakes categories it tested.

Consumers were more receptive to assistance earlier in the decision. Thirty-one percent were willing to let AI narrow choices for household supplies and 28% for personal electronics. The pattern is consistent with the shopping-task data: people want help finding information, comparing prices, surfacing deals, and reducing a long list of options while keeping final control.

Trust remains a constraint. Among consumers who had recently used AI while shopping, 54% told Gartner they had to double-check all the information the tool provided, and 62% said AI information had ended up wasting their time. AI can influence the shortlist while still creating a verification step.

05

The answer can influence a decision without producing a click

Referral traffic captures only the buying journeys in which a user selects a link. It cannot measure a business being named, compared, dismissed, or omitted inside an answer that the customer accepts without visiting a website.

Pew Research Center observed this effect inside Google. In a dataset of 68,879 unique Google searches, approximately one in five produced an AI summary. Users clicked a traditional result on 8% of visits with an AI summary, compared with 15% of visits without one. A source link inside the AI summary received a click on only 1% of visits.

Google AI summaries and ChatGPT are different products, so these click rates should not be transferred from one to the other. The relevant behavioral point is narrower: when a generated answer resolves part of the question on the results surface, influence can occur without a visit to the cited website.

06

When AI users do click, they arrive with stronger intent

Adobe Digital Insights analyzed more than one trillion visits to U.S. retail websites. During the first three months of 2026, traffic from AI sources grew 393% year over year. In March alone, it was 269% higher than one year earlier.

The quality of those visits changed even more. In March 2026, AI-referred retail traffic converted 42% better than non-AI traffic. Engagement was 12% higher, visitors spent 48% longer on the site, and they viewed 13% more pages per visit. One year earlier, AI-referred traffic had converted 38% worse, making the reversal especially notable.

This does not prove that an AI recommendation caused the higher conversion rate. People who click an AI answer may already have stronger intent, and Adobe reports on retail sites represented in its analytics data. The defensible conclusion is that AI referrals are no longer merely growing; within this dataset, they represent an unusually qualified segment of arriving traffic.

07

What the evidence means for businesses

The combined data places AI between broad discovery and the final decision. Customers use it to understand categories, express detailed constraints, compare options, obtain recommendations, and narrow choices. They may then verify the answer through Google, reviews, source links, or the business website before purchasing.

That changes where visibility begins. A business can lose consideration before a website session exists, because the generated shortlist may omit it. It can also gain influence without receiving a click, because the answer can name and frame the business directly.

The business implication is not to abandon traditional search. It is to cover the complete decision path: understand the questions customers ask, earn accurate representation in sources AI can access, and maintain official pages that let a customer or model verify the recommendation. AI search is already important because it is joining those stages together.

08

Methodology and limitations

This analysis combines separate datasets because no single source measures the full journey from chatbot adoption to purchase. OpenAI reports platform usage and conversation categories. Pew measures representative U.S. survey responses and observed Google browsing behavior. Adobe measures retail traffic represented in Adobe Analytics and supplements it with consumer surveys. Gartner measures reported shopping attitudes in smaller U.S. samples.

The percentages must not be multiplied or treated as sequential stages of one funnel. The studies use different dates, populations, products, questions, and measurement methods. Surveyed behavior can differ from observed behavior, U.S. findings may not generalize internationally, and high growth rates can begin from a small base.

The strongest conclusions are the ones repeated across methods: conversational AI has broad reach, information and decision support are major uses, shopping use concentrates around research and recommendations, many customers retain control and verify answers, and AI-referred retail visitors can represent high-intent traffic.

FAQ

Powiązane analizy