The agents helping users snag iPhone foldables are shaping the future of e-commerce in the AI era

On the morning of October 16, Apple's first foldable phone, the iPhone Duo, opened for pre-order. Starting at $1,999, with first-batch capacity tight, delivery dates soon stretched to January of next year.

This time, it wasn't just people glued to their screens racing for the first batch.

This year, Agents became the strictest father of iPhone Duo scalpers | Image source: Apple

On the MacRumors forum, a user who failed to get one vented angrily that someone on Reddit was bragging about how their configured Muse bot had snapped up several units in one go—before real people even had a chance to click, the stock had been swept away by AI.

Muse is the personal Agent Meta launched on September 8. After going live it briefly topped the US App Store free chart, with downloads quickly exceeding 2.5 million. Meta said shopping has already become one of Muse's main use cases. Right around the iPhone Duo's launch, Muse even released a new version adapted for the new device, and the line immediately following in the update notes was about making shopping in chats more convenient.

But not every store welcomes these new customers. In late September, Amazon began banning Muse. By October, users of Musk's Grok Bot also hit the same warning when placing orders on Amazon: unauthorized AI agents continuing to access violates Amazon's terms of use.

Another company, Shopify, took the completely opposite path. On September 21, Shopify CEO Tobi Lütke announced a deep partnership with Muse, letting Muse use Shop Pay to complete checkout at all Shopify stores.

And Shopify's financial reports have already shown early returns on this path. In the first quarter of this year, orders from AI search grew nearly 13-fold year over year, and store traffic brought by AI grew 8-fold.

One side closes the door, the other opens it. As more and more shopping decisions are handed to Agents, will "whether to let Agents in the door" become a key variable in e-commerce rankings in the AI era?

01 The space behind 13-fold growth

Shopify is the most active "player" in this round of the AI shopping boom.

In the fourth quarter of 2025, it reported AI-driven orders growing 15-fold. In the first quarter of 2026, store traffic brought by AI grew 8-fold year over year, orders generated by AI search grew nearly 13-fold, and new-buyer orders from AI search grew at nearly twice the rate of traditional organic search. In the same quarter, Shopify posted its fastest revenue growth in more than four years.

By the second quarter, the multiples fell back to 3-fold—both AI-driven traffic and orders were three times last year's same period. But this quarter's structure is more telling: 75% of AI-attributed purchases happened outside the platform's top 100 categories, and AI searches completed through Shopify Catalog had a conversion rate twice that of ordinary AI search.

What this set of data shows is "growing fast," not "large in scale."

Shopify itself is quite candid. On the second-quarter earnings call, President Harley Finkelstein acknowledged that agentic commerce transaction volume is still very small relative to the platform's massive GMV. To date Shopify has never disclosed the absolute value or share of GMV from AI channels, and its second-quarter GMV was $116 billion.

The Muse Agent wallet page supports Shopify's Shop Pay | Image source: Geek Park

From 15-fold to 13-fold to 3-fold, the multiples are converging rapidly—a typical curve as the small-base effect fades. The methodology is changing too: the first quarter talked about "AI search orders," the second quarter switched to "AI-driven traffic and orders," and the two sets of numbers may not add up to a continuous trend line.

Shopify's second-quarter revenue grew 34% year over year to $3.6 billion. What really supports this growth rate is GMV growth of 32%, Shopify Payments penetration rising to 68%, and Shop Pay GMV growing 53% year over year. Finkelstein also stated clearly that transactions brought by Agents are charged the same as ordinary online store transactions, with no extra fees. In other words, Agent orders' contribution to revenue can only be hidden inside GMV and payments—it can't be broken out separately.

But the significance of 13-fold isn't how much revenue it contributes today, but that it proves Agents can bring Shopify customers it otherwise couldn't get.

The first signal is new buyers. New-customer orders from AI channels grow at about twice the rate of other channels, and Shopify also took care to emphasize that traditional search remains one of merchants' largest traffic sources and is still growing. That is, AI isn't just moving traffic that originally belonged to search—it's bringing incremental traffic.

The second signal is the long tail. 75% of AI orders fall outside the top categories, which happens to be Shopify's core base. Millions of independent small shops have historically struggled to win exposure in search ads and the bidding wars on e-commerce platforms, whereas Agents are better at matching products to specific needs—and for the first time, these small shops have a chance to be "found."

To let merchants see this, Shopify added an agentic section to its admin backend in May this year, where merchants can see orders and conversions coming from ChatGPT, Copilot, Google AI Mode, Gemini, and its own Shop App. What Shopify is doing is turning the Agent from a concept into a visible sales channel in the merchant backend.

02 Amazon is blocking not Agents, but other companies' Agents

Amazon's hard line against external Agents has a long history.

In November 2025, Amazon sued Perplexity, alleging that the shopping Agent in its Comet browser entered Amazon's systems without authorization. In March this year, the court granted a preliminary injunction, and Perplexity has already appealed to the Ninth Circuit.

Muse is the second. Amazon's stated reasons are specific: Meta never informed Amazon that Muse would access the Amazon store, Muse does not identify itself when browsing, and it is suspected of scraping and storing users' account credentials. Amazon asked Meta to proactively remove Amazon from Muse, was refused, and so banned it outright.

Grok Bot is very likely the third. Amazon has not publicly confirmed it, but multiple users have posted the same block notice. Musk first suggested users have the Bot buy elsewhere, and posted that his own Bot successfully placed an order on Lego's official website, then added half-jokingly, "Maybe we should build an Amazon competitor."

Security concerns are not unfounded. Recently, a Canadian user publicly complained that Muse, without asking his consent, accepted a buyer's offer for him on Marketplace and even sent the buyer his home address. When an Agent operates everywhere using a user's account, who is responsible when something goes wrong—there is still no clear answer today.

But security is only the reason on the surface; underneath is business.

An Agent that shops for you won't see sponsored product slots, nor be swayed by homepage recommendations. Amazon's advertising revenue exceeded $56 billion in 2024, and this business is built on the premise that "people browse on Amazon." When Amazon sued Perplexity, the complaint mentioned the trouble Agent access causes for ad impression measurement.

Amazon has its own AI shopping assistant Rufus | Image source: GeekWire

Interestingly, Amazon does not oppose Agents per se.

It has its own shopping assistant Rufus, used by more than 300 million people in 2025. In May this year, Rufus and the upgraded Alexa+ merged into Alexa for Shopping, placed directly into Amazon's main search box. It is also reaching outward: the Buy for Me feature lets Amazon's AI place orders on other brands' official websites on users' behalf, and as of March this year, the underlying Shop Direct already covers more than 100 million products from over 400,000 merchants.

A more subtle step came in April this year, when Amazon joined the technical committee of UCP, the Agent transaction protocol led by Google and Shopify.

Amazon's logic is actually clear: it can send Agents to buy things at other people's houses, but other people's Agents cannot casually come to Amazon to buy things.

Shopify's choice is exactly the opposite, and the reason likewise comes from its business model.

Shopify does not have a super shelf that consumers visit to "browse" every day; it serves millions of independent stores, and its revenue comes from subscription fees and payment take rates, not ad slots. For it, every additional Agent—ChatGPT, Gemini, Copilot, Muse—means one more storefront for merchants, and there is nothing to protect.

According to Shopify, Shopify merchants are discoverable and purchasable by default in Muse, with product data synced over through Shopify Catalog. More critical is checkout. Shop Pay already stores users' bank cards and shipping addresses, so Muse doesn't need to use a user's account and password at each store to simulate operations; instead it goes through a checkout channel approved by Shopify.

For Agents that operate within the user's browser, Shopify has also prepared dedicated interfaces, providing structured tools on store pages and checkout pages so that Agents can accurately complete orders without having to parse web pages meant for humans.

Amazon sells shelves, so it must defend the foot traffic in front of those shelves; Shopify sells utilities, so it wants its pipes to reach everywhere. On the surface, the two companies' disagreement is a fight over security versus openness; in essence, it is the most rational choice for each of two different business models.

03 The emerging Agent transaction network

Muse is just the newest node on Shopify's network. In fact, Shopify is betting on three layers at once.

The bottom layer is products. Shopify has organized more than 1 billion products into clearly structured data, complete with real-time prices and inventory, which Agents can retrieve in seconds. It has also launched Agentic Plan, which lets brands that don't open stores on Shopify connect their products to Shopify Catalog and use this pipe to appear in various AI systems.

The middle layer is protocols. At the National Retail Federation conference in the US in January this year, Google and Shopify released the Universal Commerce Protocol (UCP), developed jointly by the two companies, with Etsy, Wayfair, Target and Walmart participating in its development and more than 20 retail and payment companies including Visa, Mastercard, Stripe, American Express and Best Buy lending their backing.

Full view of the UCP protocol | Image source: Google

UCP covers the entire process from product discovery and adding items to the cart to checkout and after-sales service, and it insists that merchants remain the principal party to transactions and that customer relationships belong to merchants. By April, Amazon, Meta, Microsoft, Salesforce and Stripe had joined the technical committee, expanding the number of seats from 5 to 10.

The top layer is checkout. Shop Pay has processed a cumulative transaction volume of more than $400 billion, 75% of users will use it again within 12 months, and in 2025 the average order value of Shop Pay orders was about 16% higher than that of non-Shop Pay orders. Shopify has also launched Checkout Kit, which allows AI applications to embed the entire checkout process, including Shop Pay, directly into conversational interfaces.

The counterpoint to this approach is OpenAI.

In September 2025, OpenAI teamed up with Stripe to launch the Agentic Commerce Protocol and rolled out Instant Checkout in ChatGPT, letting users place orders without leaving the conversation. It was withdrawn just half a year later. By February this year, only about 30 Shopify merchants had actually integrated Instant Checkout, and Walmart's data showed that the conversion rate for checking out directly in ChatGPT was only about one-third of that for redirecting back to the official website. OpenAI ultimately admitted that users are willing to research and compare products in ChatGPT, but are less willing to pay there.

AI companies building their own checkout counters doesn't work; Shopify's answer is to bring merchants' existing checkout counters into Agents. Users trust Shop Pay, which already stores their bank cards, and merchants trust the checkout rules they are familiar with; the Agent is only responsible for bringing people to that point.

If this path works, at least three changes will take place in the e-commerce industry in the near future.

The dimension of competition will shift from "grabbing traffic and grabbing shelf space" to "whether an Agent can understand it." Whether product data is structured, whether inventory is real-time and whether rules are clear will directly determine whether a product has a chance to appear in an Agent's recommendations. The conversion rate of Catalog retrieval is twice that of ordinary AI search, which already illustrates this point.

Wallets and checkout will become new moats. Whoever holds the user's identity, payment method and shipping address will be the most trusted handoff point in Agent transactions. Muse has integrated both Stripe's Link and Shopify's Shop Pay, and the battle at this layer has only just begun.

The e-commerce landscape will split into two routes: open alliances and walled gardens. Market forecasts cited by the institutions behind UCP hold that by 2030, Agent shopping could account for 10% to 20% of US e-commerce spending, a scale of between $190 billion and $385 billion. This figure comes from stakeholders and needs to be discounted, but the direction is not hard to judge.

The problem is that openness does not come without a cost.

Image source: Ignitiv

For merchants, when Agents make choices on behalf of consumers, carefully designed store pages and brand awareness built up over years will both be weakened, and merchants may degenerate from "stores" into a line of supply data in the eyes of an Agent. Advertising budgets that originally flowed to search and e-commerce platforms may also migrate to Agents' recommendation results. xAI previously announced that it would let advertisers pay to enter the suggestions given by Grok.

For Shopify, UCP is an open-source protocol, Stripe sits in both camps at once, and even Amazon has joined the technical committee. The protocol itself locks no one in; Shopify's real chips can only be Catalog's data and Shop Pay's users.

Of course, what the Agent transaction era brings is not all upside. For the entire industry, the hardest thing is responsibility. When an Agent places a wrong order, pays the wrong amount or leaks information, should the Agent company, the wallet, the merchant or the user bear the responsibility? Musk's approach is a public commitment that SpaceX will cover errors made by Grok Bot, but this is obviously not an industry solution that can be replicated.

Back to the launch of the iPhone Duo. In that forum post complaining about AI grabbing stock, someone raised a question: now that Agents have come into their own this year, does that make it even harder for real people to get a new phone?

This may be the most realistic preview of the Agent shopping era.

Agents are already the most diligent and tireless customers, and every e-commerce platform has to answer whether to treat them as crawlers to be guarded against or as buyers worth winning over.

Amazon and Shopify have given diametrically opposite answers, and the next round of e-commerce rankings will likely start with this multiple-choice question.

This article comes from the WeChat public account "Geek Park" (ID: geekpark), author: King of Techno, published by 36Kr with authorization.