ChatGPT Hit 1 Billion Mobile MAU, but AI Shopping Is Still Two Markets

Sensor Tower’s State of AI 2026 findings record that ChatGPT became the fastest mobile app to reach 1 billion monthly active users in May 2026, while generative-AI referrals increased across every major retail category from Q4 2024 to Q1 2026 and Amazon shoppers using Rufus converted at nearly twice the rate of non-users. The audience milestone remains intact, but the commerce evidence now supports a more qualified conclusion: AI shopping is developing through two distinct channels rather than becoming one uniformly powerful source of retail traffic.

The clearest subsequent change occurred inside Amazon, where Rufus was incorporated into a more capable service called Alexa for Shopping. That expansion strengthens the case for retailer-owned assistants, while independent transaction evidence still places external ChatGPT referrals in a small, unevenly performing segment of e-commerce.

Audience reach is not the same as retail influence

ChatGPT’s billion-user milestone matters because it gives conversational product discovery enormous potential distribution. A consumer can ask for comparisons, narrow a purchase by specific constraints and receive links without beginning with a conventional search engine or retailer category page.

Monthly activity nevertheless measures access to the app, not purchase intent. It does not establish how many conversations concern products, how frequently users follow retail links or how much revenue those visits generate. App reach, referral traffic and completed transactions describe different stages of a commercial journey and cannot substitute for one another.

The same caution applies to the report’s comparison between Rufus users and other Amazon shoppers. Rufus operated inside a store where visitors could already be signed in, browsing products or close to buying. Higher conversion among assistant users may reflect the value of the feature, greater initial intent or both; the comparison alone does not isolate causation.

Amazon has turned the embedded assistant into a shopping layer

On May 13, 2026, Amazon’s Alexa for Shopping announcement presented the service as a combination of Rufus and Alexa+, noted that Rufus had helped more than 300 million customers during 2025, and made the successor available to signed-in US customers through the Amazon Shopping app and website without requiring Prime or an Echo device. The same page describes product comparisons, shopping guides, cart building, recurring actions and access to as much as one year of price history.

This is more than a label change. The assistant can participate across discovery, evaluation and cart creation because it is connected to Amazon’s catalog, account context and transaction environment. For eligible products offered elsewhere, its Buy for Me capability can also complete a purchase from another store.

An embedded assistant therefore has structural advantages that an external chatbot usually lacks. It can remain present as a shopper moves from a broad question to individual products, prices and an order, while the retailer can observe more of that journey within its own systems.

That distinction explains why strong engagement inside Amazon does not prove that chatbot referrals will deliver the same performance on independent retail sites. The user context, data access, interface and point in the purchase journey are materially different.

External ChatGPT traffic remains a developing channel

A peer-reviewed Marketing Science study published online on April 21, 2026 analyzed 12 months of first-party data from 973 e-commerce websites representing $20 billion in combined annual revenue, including more than 50,000 ChatGPT-referred transactions and 164 million transactions assigned to traditional channels. Its scale supplies a broader counterweight to results drawn from one retailer’s embedded feature.

Within that dataset, organic ChatGPT referrals delivered conversion rates and revenue per session above paid social but below every other traditional channel in the comparison. Results were stronger in complex product categories, where conversational research may help shoppers reconcile more attributes, yet overall traffic volume and revenue per session remained modest.

Engagement was mixed rather than uniformly superior. ChatGPT-referred visits had favorable bounce behavior but shorter sessions and fewer page views, suggesting that some users arrived with a narrower informational task. The channel consequently looked useful for particular research-heavy journeys, not like a general replacement for search, email, affiliates or direct traffic.

The comparison also has an attribution boundary. Last-click measurement can miss an assistant that shapes early research but does not generate the final visit. Conversely, a high conversion rate among a small audience can appear impressive while producing little aggregate revenue, so conversion quality and total contribution must be considered separately.

The commercial shift is real, but it is happening at different speeds

Sensor Tower’s central observation survives scrutiny: conversational systems are becoming part of product discovery, and the effect is especially plausible where buyers face complicated choices. What changes after incorporating the newer evidence is the definition of that shift.

Retailer-owned assistants currently have the deeper commercial role. They can use catalog structure, customer context and transaction controls to move beyond recommendations into comparison, price monitoring and cart actions. Their value resembles a product-experience improvement inside the store as much as a new marketing channel.

External assistants occupy a different position. They can introduce retailers to high-intent shoppers and influence decisions before a site visit, but measured referral volume remains limited and performance varies by category and metric. Their strategic importance may therefore exceed their directly attributed revenue without making the channel large today.

The billion-user headline and the niche-referral finding are not contradictory. One describes the reach of a general-purpose mobile assistant; the other describes a subset of visits leaving that assistant for retail websites. The more consequential e-commerce story is the divergence between those external referrals and assistants embedded directly in stores.

For businesses, the distinction determines what can reasonably be inferred from the data. ChatGPT’s scale creates a large surface for product research, but it does not guarantee retailer traffic or sales. The stronger near-term evidence lies in assistants that remain inside a commerce environment long enough to help a shopper compare, decide and transact.

Also read:

• Uzbekistan’s Uzum Hits $2.3 Billion Valuation in Record Time, Raising $131.5 Million and Becoming the Country’s First Unicorn
• Microdramas? No, Microseries. Vertical Video Industry Revenue Set to Hit $150 Billion in 2026 (Ex-China)
• MrBeast Just Hit 500 Million Subscribers — And He’s Still the Only One