Consumer AI usage and competitive adoption: ChatGPT remains the consumer-AI leader, but Gemini, Claude and Meta’s Muse are reshaping competitive momentum.

Overview

This recurring Bernstein tracker examines consumer-AI adoption, engagement, web traffic and the emergence of agentic products. It concludes that ChatGPT remains the category leader, but competitive share is shifting toward Gemini and Claude, while Meta’s Muse offers an early test of whether consumer AI can progress from answering questions to completing tasks.

Core views

Bernstein tracks downloads, monthly and daily active users, retention, app engagement and web traffic to assess consumer-AI adoption and competitive momentum. ChatGPT remains the clear leader globally and in the United States: it accounts for roughly 64% of global AI-app MAUs and about 60% in the U.S., with a still larger lead in DAU because its DAU/MAU ratio is about 40–45% and its seven-day retention is about 25–30%. ChatGPT has also surpassed 1 billion weekly active users and remains the leading standalone AI application. The report views these usage measures as evidence that ChatGPT is more frequently and consistently used than major peers. The leadership position is being challenged. ChatGPT download growth turned negative year on year in both global and U.S. markets beginning in April 2026. Globally, its August downloads declined 27% year on year, versus growth of 79% for Gemini, 754% for Claude and 363% for Meta AI; combined downloads across the five tracked chatbots rose only 6%, which Bernstein sees as an early sign of plateauing. In the U.S., ChatGPT downloads fell 9% in August while Gemini, Claude and Meta AI grew 29%, 421% and 471%, respectively; combined downloads grew 18%. Bernstein notes that Gemini’s app-download share may understate newly activated users because it is pre-installed on some Android devices. MAU data show both continued category expansion and a redistribution of incremental users. In August, global MAUs grew 70% year on year for ChatGPT, 190% for Gemini, 1,228% for Claude and 704% for Meta AI; the combined five-app cohort grew 108%. U.S. MAU growth was 49% for ChatGPT, 185% for Gemini, 1,031% for Claude and 611% for Meta AI, with the combined cohort up 95%. ChatGPT’s global MAU share fell from 78% in August 2025 to 64% in August 2026, though it still captured about 55% of incremental MAU growth from May through August 2026. Gemini captured 36% of global incremental MAU growth over that period and lifted its share from 19% to 26%; Claude captured about 12%. The U.S. shift was sharper: ChatGPT’s share declined from 79% to 60% and it captured about 44% of incremental MAUs, while Gemini captured 32%, rising from 13% to 19% share, and Claude captured 18% and reached a 9% share. Across platforms, engagement is becoming more standardized, which Bernstein interprets as evidence that chatbots are becoming habitual consumer products. Typical usage is roughly 14–16 minutes per day globally and 15–17 minutes in the U.S., around eight to nine sessions per day, and approximately 1.5–2 minutes per session. ChatGPT remains the engagement and retention leader; Claude has improved its retention recently, while Meta AI has had lower usage but has been narrowing the engagement gap. On the web, ChatGPT remains the leader in global and U.S. traffic, although its share has steadily ceded ground to Gemini and Claude before recently stabilizing at roughly 60%; ChatGPT.com ranks No. 5 globally by traffic. Most chatbot sites have average visit duration near six minutes, versus about 12 minutes for Grok, which the report attributes potentially to a narrower but more engaged and technically savvy user base. The report also identifies AI as an emerging, but still small, consumer-discovery channel. AI referrals contribute less than 1.5% of overall traffic to e-commerce and retail sites. Nonetheless, Booking has acknowledged AI effects on portions of its search-driven discovery funnel; Reddit’s CEO said Google AI Overviews had not been a meaningful traffic source, raising concern that AI-generated answers could reduce referrals to open-web destinations. Bernstein also cites weaker online lead generation and reduced organic search traffic at Rollins’ Orkin business as a possible analogous development. The firm concludes that the effects are uneven and difficult to quantify, but AI is increasingly becoming an intermediary between consumers and businesses. Meta’s September 8, 2026 launch of Muse is the report’s principal agentic-AI case study. Within 10 days, Muse climbed to No. 1 in the U.S. App Store and overtook ChatGPT on September 17; downloads after launch were already trending above those observed after ChatGPT’s May 18, 2023 app launch. Bernstein attributes the early traction partly to Meta’s distribution through Instagram, Facebook and Threads, as well as referral incentives that grant additional usage credits. Consumer checks found its conversational, personalized interface encouraging, and Bernstein’s testing found that Muse could reasonably handle tasks such as doctor appointments, reminders, personal goals, flight-price monitoring, customer-service contact and subscription cancellation. Its social context from Instagram and Threads may aid trip planning, local discovery and restaurant recommendations. However, Bernstein stresses that download momentum alone does not establish a durable product. Muse performed inconsistently in some seemingly straightforward tasks, including hotel reservations. Privacy concerns may limit permission grants, narrowing functionality; consumers may also prefer to consolidate AI use around ChatGPT, Claude or Gemini rather than add another assistant. The firm expects investor attention to move toward seven- and 28-day retention, WAU/MAU expansion, DAUs, DAU/WAU and DAU/MAU as indicators of routine utility. Meta currently offers a free plan with up to 100M tokens per week, a $20-per-month Power tier with 500M tokens per week, and a $100-per-month Maximum tier with 3B tokens per week. Bernstein expects most users initially to remain on the free plan, implying limited near-term monetization as Meta prioritizes adoption, engagement and product iteration while subsidizing inference costs. Finally, Bernstein frames the agentic competition as fluid. Amazon’s Alexa+ is supported by a base of more than 600 million devices; more than 350 million customers used Alexa in the prior 12 months, interactions rose fivefold year on year, and Alexa+ users engaged two to three times more frequently. Google’s Gemini benefits from distribution across Search, Maps, Gmail and Android, with the Gemini app above 1B MAUs, AI Mode above 1B users and AI Overviews above 2.5B monthly users. Bernstein highlights Gemini Spark and Daily Brief as building blocks for a proactive personal agent. It expects continuing launches, partnerships and announcements to rotate market attention quickly, and notes that an updated OpenAI consumer-agent offering may be forthcoming.

Analysis framework

Bernstein compares major consumer-AI platforms using app downloads, MAU and DAU trends, user-share and incremental-user capture, retention, session behavior and web-traffic data. It then connects those measures to platform distribution strategies, tests Meta’s Muse through consumer checks and initial product testing, and evaluates whether emerging agent products show evidence of sustained daily utility rather than one-time download interest.

Methodology notes

• Industry AnalysisIndustry Concentration Analysis

  Platform-share and incremental-user analysis

  The report compares MAU and web-traffic shares and the proportion of new users captured by each chatbot to show how competitive leadership and concentration are changing.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  AI-mediated consumer discovery and referral traffic

  The report examines how AI assistants and AI search can sit between consumers and online businesses, potentially changing search discovery and referral traffic to commerce and local-service destinations.

• Other

  Engagement-cohort tracking

  The report uses DAU/MAU, seven-day retention, time spent, sessions and session duration to distinguish sustained routine usage from superficial user growth.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Alphabet (GOOGL)

  Covered company with Gemini benefiting from broad distribution across Google consumer products.

  Strengths

  Gemini app surpassed 1B MAUs; AI Mode exceeded 1B users and AI Overviews exceeded 2.5B monthly users.

  Weaknesses

  The report notes investor focus on the absence of a significant recent flagship-model upgrade relative to faster competitor release cadence.

  Comparison

  Gemini is the clear No. 2 globally and gained MAU share while ChatGPT lost share.

• Meta Platforms (META)

  Covered company whose Meta AI and Muse products are central to the report’s agentic-AI discussion.

  Strengths

  Meta AI surpassed 1B MAUs; Muse gained rapid early downloads supported by Meta’s app ecosystem and social context.

  Weaknesses

  Muse task reliability is inconsistent in some use cases, and near-term monetization is likely limited.

  Comparison

  Muse reached No. 1 in the U.S. App Store within 10 days and overtook ChatGPT on September 17.

  Risks

  Privacy concerns, restricted permissions and weak sustained engagement could limit adoption.

• Amazon (AMZN)

  Covered company linked to agentic commerce through Alexa+.

  Strengths

  More than 600M-device distribution network; Alexa interactions rose 5x YoY and Alexa+ users engage 2–3x more frequently.

  Comparison

  Amazon’s shopping intent, payments, fulfillment and distribution differentiate its potential agentic-commerce position.

• Reddit (RDDT)

  Covered company cited as an open-web destination potentially affected by AI search.

  Weaknesses

  Google AI Overviews were not a meaningful traffic source, according to Reddit’s CEO.

  Comparison

  The report contrasts AI-led discovery with conventional referral traffic to open-web destinations.

  Risks

  AI-generated answers may reduce referrals to open-web destinations.

• Rollins (ROL)

  Example of a local-services company potentially exposed to changing AI-led discovery.

  Weaknesses

  Orkin reported weaker online lead generation amid a sharp decline in organic search traffic.

  Comparison

  Cited alongside Booking and Reddit as evidence that AI-discovery effects may extend beyond commerce.

  Risks

  Large pest-control brands may be less frequently recommended by certain AI assistants.

Key data

• ChatGPT global AI-app MAU share~64%Current global share; down from 78% in August 2025.
• ChatGPT U.S. AI-app MAU share~60%Down from 79% over the prior year.
• Global August download growthChatGPT -27% YoY; Gemini +79%; Claude +754%; Meta AI +363%Combined downloads across five tracked chatbots were +6% YoY.
• U.S. August MAU growthChatGPT +49% YoY; Gemini +185%; Claude +1,031%; Meta AI +611%Combined MAUs across the five apps rose 95% YoY.
• Global August MAU growthChatGPT +70% YoY; Gemini +190%; Claude +1,228%; Meta AI +704%Combined MAUs across the five apps rose 108% YoY.
• ChatGPT engagement~40–45% DAU/MAU; ~25–30% seven-day retentionHighest among major chatbot platforms according to the report.
• Typical chatbot engagement~14–16 minutes/day globally; ~8–9 sessions/day; ~1.5–2 minutes/sessionU.S. daily usage is about 15–17 minutes.
• AI commerce referral traffic<1.5% of total trafficApplies to e-commerce and retail websites; the channel remains early-stage.
• Muse launch outcomeNo. 1 in U.S. App Store within 10 daysLaunched September 8, 2026 and overtook ChatGPT on September 17.

Impact & implications

Bernstein sees consumer AI shifting from a scale race toward a competition for habitual engagement and agentic utility. ChatGPT’s scale and engagement remain important advantages, but challenger share gains and Meta’s Muse launch suggest that distribution, ecosystem integration, reliable task completion and retention will increasingly determine competitive outcomes. AI’s influence on commerce discovery may become material over time, although current referral traffic remains small.

Risks

• Muse’s early downloads may not translate into sustained engagement, retention or adoption.
• Muse has shown inconsistent performance in some agentic use cases, including hotel reservations.
• Privacy concerns may cause users to restrict permissions, materially narrowing agent functionality.
• Consumers may consolidate around an existing primary assistant rather than adopt another AI product.
• AI’s effect on referral traffic and consumer discovery remains uneven and difficult to quantify.
• Rapid product launches and new entrants could quickly change market narratives and competitive attention.

What to watch

• Muse’s seven- and 28-day retention, WAU/MAU growth, DAUs, DAU/WAU and DAU/MAU.
• Whether Meta’s early Muse downloads convert into sustained daily utility and engagement.
• Gemini and Claude share gains, especially in U.S. MAUs and incremental-user capture.
• ChatGPT download trends and whether its roughly 60% web-traffic share remains stable.
• The evolution of AI referral traffic and impacts on discovery funnels, open-web referrals and local-service lead generation.
• New agentic offerings from Google and OpenAI, and continuing product launches across major AI platforms.