Consumer AI Adoption Stalls at 64% as Spending Triples to $40B

Consumer AI's growth now comes from heavier use, not new users

For three years, consumer AI was measured by how many new people tried it. The latest 2026 data points to a different measure. The growth now comes from people who already use AI using it more, paying more for it and handing it more control.

Menlo Ventures' 2026 State of Consumer AI report makes this case most directly. The share of U.S. adults using AI rose only from 61% to 64% over the year, while estimated global consumer spending more than tripled to about $40 billion.11 The firm estimates that the global user base grew about 11%, from 1.8 billion to 2.0 billion people, while spending rose from $12 billion to $40 billion.11 Daily use among all Americans climbed from 19% to 25%, and 55% of AI users now pay for at least one AI product.11

The market is still growing, but in a different way. Consumer AI now looks less like a frontier being settled and more like a maturing subscription economy. A loyal core spends more each year, and a group of holdouts is becoming more opposed.

ChatGPT keeps growing while its share shrinks

On raw numbers, ChatGPT keeps setting records. OpenAI reported 900 million weekly active users in February 2026, up from 400 million a year earlier.24 Several trackers say it passed 1 billion weekly users in August.2528 A late-September developer event put the figure at 1.2 billion.27 Reports also credit the ChatGPT app with being the fastest app ever to reach 1 billion monthly active users. One tracker dates that to June22, while another, citing Sensor Tower, says May.28 The company reports more than 50 million paying subscribers across its consumer tiers.21

Competitive data tells a different story. Sensor Tower found that ChatGPT's share of the global AI assistant app audience fell below 50% for the first time, to 46.4% at the end of May 2026. Gemini held 27.7% and Claude 10.3%.56 Similarweb-based web-traffic figures show the same trend. ChatGPT's worldwide share of visits fell from 76.5% in early 2025 to 53.9% in May 2026, while Gemini rose from 5.6% to 27.9%.2 That analysis also found ChatGPT's visit count roughly flat while competitors took nearly all of the market's growth. It called this erosion from a plateau, not a decline in use.2

Survey data shows the same narrowing. Menlo found that 60% of AI users use ChatGPT, but Gemini jumped from 36% to 58%, cutting ChatGPT's lead from nine points to two.11 Andreessen Horowitz's latest consumer rankings reportedly show ChatGPT's U.S. paid-subscriber lead over Claude shrinking from eight times to about three times in roughly seven months.16

This is where the measurements disagree most. StatCounter's referral-based figures still gave ChatGPT nearly 77% of web traffic share in June, far above Sensor Tower's or Similarweb's numbers.6 Platforms also count users differently. Meta AI's reported 1.2 billion monthly users include people who encountered it passively inside WhatsApp, Instagram or Facebook.1 Other estimates put Meta AI at about 500 million.4 No single figure decides the race. Still, every dataset that tracks change over time moves the same way, and that makes the narrowing trend credible even where the exact numbers are not.

Why the field is opening up

The biggest structural change may be that consumers no longer pick just one assistant. Menlo found the average AI user now uses 3.0 general-purpose assistants, up from 2.2 a year ago. Combined reach across ChatGPT, Gemini, Claude and Perplexity now exceeds 150% of users.11 When most people use several assistants, ChatGPT can lose share without losing users.

Gemini's rise comes mainly from distribution. Sensor Tower credits its growth to Android integration and Google's wider ecosystem, especially in Europe, the U.S., Japan and South Korea.17 Google says the Gemini app passed 1 billion monthly users in August, up from 400 million 15 months earlier.3 One analysis points to a deal for Gemini to power Siri as the year's biggest distribution event.1

Claude's growth came in bursts. Its share of downloads among leading AI apps jumped from about 1% in every quarter of 2025 to 14% in the second quarter of 2026.9 Several outlets connect a February spike to Anthropic's statement that it would not partner with the Pentagon. After that, Claude topped the U.S. App Store and beat ChatGPT in daily U.S. downloads for the first time.510 The momentum did not last, however. Card-panel data reportedly showed Claude's daily sessions falling in July and August, while ChatGPT sped up again after new model releases and the launch of ChatGPT Work.16 Only 8% of U.S. ChatGPT subscribers also pay for Claude.16

Changes to the ChatGPT app

OpenAI's product decisions are part of this story. The company has added a cheaper tier, ChatGPT Go, at $8 a month, alongside Plus at $20 and Pro at $200.21 Andreessen Horowitz's Olivia Moore reportedly sees cheaper options like this as key to widening a market where subscriptions are the main bottleneck.16 The flagship model was updated in July with the GPT-5.6 family.26

Ads are the more sensitive change. Sensor Tower data shows ChatGPT's churn rate rising from 12.7% in January to 14.5% in April, the same period ads arrived on the platform.8 That timing does not prove the ads caused the churn. It does suggest the new revenue model may cost the company some users.

Engagement still favors OpenAI. Sensor Tower measured about 215 minutes per ChatGPT user per month, compared with 120 minutes for Claude and 100 for Gemini, though both challengers are rising quickly.17 Global time spent in generative AI apps was projected to reach 36 billion hours in the first half of 2026, more than double the year before.17

A small group of power users drives spending

The spending growth is heavily concentrated. Menlo found that the top 14% of payers account for 60% of consumer AI spending. Payers are also about twice as likely as non-payers to use AI daily and five times as likely to use AI agents.11 Card data cited by Andreessen Horowitz shows a similar pattern. The top 1% of spenders accounted for 19.5% of observed U.S. consumer AI spending in August and averaged $903 a month, mostly on automation, building and creative tools.16 Moore's view, as reported, is that much of what gets called consumer AI is really prosumer AI.16

This is the most important disagreement in the data. Menlo says 55% of AI users pay for something.11 Bank of America's transaction data found only about 3% of its households paying for AI services as of March, although that number was rising quickly.19 Part of the gap comes from method: survey answers and bank-card records measure different things. Still, the difference is large enough to make some survey-based estimates of willingness to pay look optimistic. A careful reading is that paid AI is real and growing but probably narrower than the headline survey numbers suggest.

Agents and the question of trust

AI agents are where consumers are handing over more control. Menlo reports that 41% of AI users have tried an agent, 24% use one regularly, and 32% have let AI act for them without final approval.11 Shopping data points to more caution. In Stord's survey, 30% of consumers said they would never let AI handle purchases or payment details. Only 16% were very comfortable with it.14

Trust is now what decides how far people go with AI. Menlo found that users rank accuracy, trustworthiness and privacy above ease of use. Among non-users, distrust of AI-generated information rose from 58% to 70%.11 Prophet's five-country study found GenAI use rising to 73%, while excitement fell about 7% and belief that AI will handle most decisions dropped 30%.13 Mintel puts U.S. use at 68% and finds ethical concerns highest among younger adults.20 There is also resistance to AI-made content: 36% of consumers say they are less likely to engage with content they know is AI-generated.11

Outlook

The data supports a clear conclusion. Consumer AI is no longer an adoption story. It is a story about monetization and trust. ChatGPT is still by far the largest single product, but it now competes in a market where users run several assistants, switch quickly and increasingly choose based on reliability and values. The companies most likely to gain are those that can turn heavy users into dependable revenue without losing the trust of the much larger group of casual users and skeptics.