Consumers Stop Sampling AI and Start Relying on It

The artificial intelligence story used to center on how many people have tried ChatGPT or asked Siri a question (sometimes about whether algorithms and robots will replace their jobs). Now it’s about how deeply the tools of artificial intelligence are embedding themselves in the most routine moments of daily life, from what people buy and how they budget to where they shop and the financial decisions they make.

The latest installment of the Agentic AI Report Series shows that AI platforms are entering a new phase, with overall adoption stabilizing at roughly half of the U.S. population. However, usage patterns reveal that what we call mainstream and power users are increasingly replacing traditional methods and relying on native AI platforms to shop for products and even make financial decisions. The data shows that growth is no longer being driven by new user acquisition but by intensifying engagement among existing users. And many of these users are treating conversational AI platforms not as a supplemental tool but as a primary interface for commerce and household financial management.

These are just some of the findings detailed in “Beyond the Try: Native AI Platforms Begin Replacing Legacy Paths to Shopping Discovery,” a PYMNTS Intelligence exclusive report. This study examines how consumers use AI when shopping and draws on insights from a survey of 2,439 U.S. adult consumers conducted from December 9, 2025, to January 5, 2026.

• AI Adoption Enters a Stabilization Phase
• The Next Phase of AI is Depth, Not Reach
• ChatGPT is the Universal On-Ramp
• Platform Experimentation is Concentrated Among Power Users
• Read More
• Methodology

AI Adoption Enters a Stabilization Phase

‘Power users’ of AI boosted their use of the technology to 27 activities last December from 25 three months earlier, while usage by other users is flat.

By December 2025, 49% of consumers and two in three millennials had used a conversational AI assistant.¹ What’s notable is not the headline number but its consistency over time. Between last October and December, usage rates remained largely flat across demographic groups, suggesting that people willing to experiment with conversational AI are already active, and that incremental growth is slowing. Strong penetration levels indicate mainstream awareness but also hint at potential saturation among early and mid-level adopters.

The data shows a clear division. Power users, mainstream users and light users have maintained relatively stable shares month over month. Meanwhile, hold-outs—those who haven’t used AI and cite specific reasons for staying away—remain unmoved.² The absence of a sharp uptake across demographics points to the adoption horizon now encountering more skeptical or reluctant consumers, for whom concerns around trust, usefulness or search habit change may outweigh perceived benefits.

When we look at the number of activities each group performs, the picture becomes clearer. Light users performed roughly two activities on average last December, the same as in September, and nearly all were low-complexity activities. Mainstream users also held steady, averaging eight activities with minimal variation in the types of tasks they performed. Power users, by contrast, increased their total activity count from an average of 25 last September to 27 in December.

Notably, this growth came entirely from increases in low and medium-complexity tasks, such as creating a shopping list or planning a travel itinerary, not high-complexity ones, such as buying online using one-click checkout or drawing up a household budget. Power users are expanding the breadth of what they use AI for, but they are not yet moving dramatically to use AI more frequently for complex or high-stakes purposes..

The Next Phase of AI Is Depth, Not Reach

One in three ‘mainstream’ users of AI have swapped their old approaches to shopping with AI-fueled alternatives.

Among existing adopters, conversational AI is increasingly displacing traditional ways of doing things, rather than complementing them. In shopping discovery-related tasks, power users were 31% more likely last December than in November to report replacing their previous search methods, signaling a shift from experimentation to reliance. Mainstream users followed a similar trajectory, with replacement rising from 22% to 30% month over month. Light users remained flat at roughly 11%.

This substitution effect extends into shopping and purchasing. Nearly half of power users, or 48%, now report having supplanted their old approaches with AI-driven alternatives, up sharply from 39% the prior month. Mainstream users showed the same pattern, with replacement climbing from 22% to 32%. For managing finances and banking, the numbers are equally striking. By December, 37% of power users reported using native AI platforms as their primary tool for these tasks. The share among mainstream users doubled, climbing from 14% to 28%.

What this reveals is that growth among adopters is driven less by new users coming on board and more by a deeper integration of AI into the everyday workflows of people already using it. Power users aren’t just using AI more frequently; they’re making it the default. Light users, by contrast, continue to treat conversational AI as an occasional or supplementary tool rather than as a primary channel.

A further signal of deepening behavioral change is the growing share of users who now treat native AI platforms—such as ChatGPT, Google Gemini or Perplexity—as their primary channel for completing tasks. In shopping discovery-related activities, 34% of power users relied on native AI interfaces as their most-used method in December, up sharply from 22% in November. Even light users are beginning to shift behavior, with reliance rising from just 5% to 16% in the same period.

ChatGPT Is the Universal On-Ramp

More than four in five adult Americans have used ChatGPT at least once.

The native AI platform landscape is highly concentrated. More than four in five users, or 83%, report having used ChatGPT at least once, far surpassing Google Gemini at 48% and Microsoft Copilot at 30%. This gap underscores ChatGPT’s role as the default entry point for conversational AI, with Gemini and Copilot remaining secondary options with lower reach.

Within smartphone-embedded conversational AI, consumer usage is converging around a small set of leading assistants. ChatGPT and Google Gemini are effectively tied as the most-used options, with 40% of users reporting at least one use of each. At the same time, 37% of users have used Google Assistant, reflecting its legacy default placement in the Google suite of products rather than an active preference. With Google Assistant set to be retired and fully replaced by Gemini by March 2026, this overlap suggests a near-term migration effect rather than fragmentation.

These patterns suggest that power users aren’t merely heavier AI users overall, but are also more experimental. They’re actively testing multiple platforms to optimize their work on different tasks, while mainstream and light users tend to consolidate around a single, familiar solution rather than engage in platform comparisons. Power users are samplers. If they need a specific capability for a specific task and believe Perplexity or Claude will deliver it better, they will go there. Mainstream and light users, by contrast, are consolidators who default to what they know.

Platform Experimentation Is Concentrated Among Power Users

The data reveals a clear split between a universal entry point and exploratory behavior. ChatGPT serves as the on-ramp, with near-total reach across all user types. But beyond that initial point of contact, behavior diverges sharply. Power users aren’t just fans of ChatGPT. They’re also using Google Gemini at rates nearly double those of mainstream users and more than double those of light users. They’re also using Meta AI, Microsoft Copilot, Perplexity, Grok, DeepSeek and Claude at rates far exceeding those of other groups.

These findings signal that power users treat AI platforms as tools, not as destinations. They go where the capability is best for the task at hand. If they need a web search integrated into their query, they might use Perplexity. When they’re looking for coding assistance, they might use Claude. If they’re embedded in the Microsoft ecosystem, they might use Copilot. This behavior is fundamentally different from that of mainstream and light users, who seek simplicity and consistency rather than variety and specialization.

The implication for AI-native platforms jostling for dominance in a fiercely competitive, fast-evolving environment is clear. Winning among power users requires differentiation on capability, not just accessibility. Winning among mainstream and light users requires trust, familiarity and seamless integration into existing workflows.

Read More

PYMNTS Intelligence is the leading provider of information on the trends driving [topic area]. To stay up to date, subscribe to our newsletters and read our in-depth reports.

• More Than 60% of Consumers Now Start Daily Tasks With AI
• The First Chatbot Consumers Try May Be the One They Stick With
• ChatGPT Leads Consumer AI Use as OpenAI Pushes New Paid Tiers

Methodology

“Beyond the Try: Native AI Platforms Begin Replacing Legacy Paths to Shopping Discovery” is based on a U.S. Census-balanced survey of 2,439 U.S. adult consumers conducted from December 9, 2025, to January 5, 2026. The report examines which consumers use AI, for what purposes and how frequently.

^{1. PYMNTS Intelligence uses the following approximate birth dates and approximate age ranges in 2026 for generational cohorts: baby boomers: born in 1964 or earlier and now aged 62 or older; Generation X: born between 1965 and 1980 and now aged 46–61; millennials: born between 1981 and 1996 and now aged 30–45; bridge millennials: born between 1978 and 1988 and now aged 38–48; zillennials: born between 1991 and 1999 and now aged 26–35; and Generation Z: born in 1997 or later and now aged 29 or younger.↩}

^{2. PYMNTS Intelligence defines power users as using AI to perform 27 or more distinct tasks each month, including higher-complexity ones such as personal investment management. They represent 10% of all consumers and 19% of millennials. Mainstream users, representing 27% of all consumers and 41% of Gen Z, perform an average of eight tasks per month, most low complexity, like comparison shopping. Light consumers, or 10% of all consumes and 12% of Gen X, perform an average of two tasks a month. Nearly six in 10 use AI to complement their traditional search methods. Holdouts represent 53% of all consumers and 72% of boomers.↩}