Key Takeaways
- The Current's analysis argues that even as consumers use AI to discover products, they still press the buy button themselves
- With cart abandonment still above 70%, the upside from better checkout is large, so the contest is decided at conversion rather than discovery
- E-commerce and booking businesses need to stop handing customer relationships to third-party AI and take back control of their own checkout and every touchpoint
AI Handles Discovery, but the Purchase Stays in the Consumer's Hands
Consumers may use AI to discover products, but retailers still need to win them at checkout.
www.thecurrent.comOn June 30, 2026, the advertising and commerce publication The Current released the first installment of "The AI Effect," a three-part series tracking how AI reshapes consumer buying behavior. The premise of Part 1 is clear. The competition in AI commerce will not be settled at the discovery stage but at checkout, the moment a purchase is actually completed.
The piece opens with a telling episode. In March this year, Amazon won a court order blocking Perplexity's AI shopping agent from accessing its website. Yet three months later, the company loudly unveiled its own agentic shopping tools, declaring in a blog post that "agentic shopping is here." At Cannes Lions, it went further with "Alexa+ Agentic Ads," which let users complete a purchase without leaving the ad unit. Keep rival agents out, but keep the moment of purchase inside your own house. That seemingly contradictory posture is exactly what confirms checkout as the main battleground.
Consumers Have Not Handed Their Wallets to AI Yet
The data The Current returns to again and again is that consumers use AI to research but not yet to buy.
An April 2026 report from Epsilon found that the leading use cases for AI-assisted shopping all clustered around research and discovery: comparing prices, doing initial research, and searching for deals. The generation most fluent with AI is Gen Z at 86%, yet even they mainly use it to gather product information and compare prices. Rachel Cascisa, Epsilon's vice president of platform adoption, put it plainly.
A lot of people are asking us about agents buying on behalf of consumers, but that isn't really happening yet. Consumers still want to make purchases themselves; they're not willing to let AI spend their dollars.Source: Rachel Cascisa, Epsilon
Other evidence points the same way. A survey of 3,000 people across the US and UK by The Trade Desk found that 88% are not fully comfortable using AI tools for big decisions, and that consumers are 1.6 times more likely to finalize a purchase on the open web outside AI tools. A Reddit survey similarly found half of US shoppers verify AI recommendations on the platform before buying. Anthony Katsur, CEO of IAB Tech Lab, noted that behavior change is hard and that "we'll be in a world of AI-assisted commerce for a while." EMarketer CEO Matthias Braun added that shopping behavior on LLMs is still limited, with AI users preferring to buy through traditional search or an app.
The upshot is that while AI is quietly seizing the top of the funnel, the final act of buying still lives in the merchant's checkout. That is precisely why businesses need to meet consumers reliably at the touchpoints where they are not using an agent, above all at the moment of purchase.
The Upside at Checkout Is Still Enormous
Why is checkout the decisive front? One reason is the sheer size of the opportunity that remains there.
According to a Baymard Institute meta-analysis of 50 studies, the average cart abandonment rate in 2026 stands at 70.22%. Roughly seven in ten shoppers who add an item to the cart never complete the purchase. Baymard estimates that improving checkout design alone could lift conversion at large e-commerce sites by an average of 35.26%, with $260 billion recoverable across the US and EU. Among the reasons for abandonment, 48% cite unexpected shipping, tax, or fees pushing the total higher than expected, and 18% point to a checkout that is too long or complicated.
What these numbers mean is that even if discovery is partly conceded to AI, there is still an enormous pool of demand to be captured by reducing friction at purchase completion. Getting lost in the fight over who owns discovery while spilling the carts already in front of you would be a strategic own goal.
There is an intriguing signal that the quality of AI-referred traffic is actually improving. Adobe's Q1 2026 data showed AI referral traffic converting 42% better than non-AI traffic. Given that a year earlier, in March 2026, it converted 38% worse, that is roughly an 80-point reversal. AI increasingly appears to function as a pre-qualification layer, sending "pre-vetted" visitors who have already done their research inside the chat. Shoppers who leave discovery to AI tend to arrive with higher purchase intent, which makes it all the more important to design a checkout that does not squander that intent.
Do Not Hand the Customer Relationship to Third-Party AI
The Current emphasizes a second question: who ends up owning the customer relationship.
Real-world cases show that partnering with external AI platforms is a double-edged sword. This year Walmart reworked its OpenAI partnership after disappointing sales, dropping the approach of letting users order directly from ChatGPT and instead integrating its own shopping assistant, Sparky, into its app and Google's Gemini. Target went as far as warning users after a Google Gemini integration that they would be on the hook for any order an AI agent mistakenly placed. Outsourcing the buying experience to third-party AI leaves both liability and the customer relationship dangling. A Commercetools white paper argues that owned experiences are precisely where agentic AI can strengthen relationships, by guiding complex decisions, resolving service issues proactively, and reinforcing loyalty.
Payment infrastructure is reinforcing this dynamic from the transaction side. The Agentic Commerce Protocol (ACP), codeveloped by Stripe and OpenAI, emerged as the standard behind Instant Checkout, which lets shoppers complete purchases of Etsy and Shopify products inside ChatGPT. Google, together with more than 60 partners including Mastercard and PayPal, introduced the Agent Payments Protocol (AP2), while Visa and Mastercard run their own tokenization schemes for agents. Each is a mechanism for moving that last step, purchase completion, smoothly through an agent. The very fact that the industry is investing so heavily in standardizing checkout tells you where the outcome is decided.
How E-commerce and Booking Businesses Can Reclaim Their Checkout
So what should e-commerce merchants and travel or booking intermediaries actually do at this juncture?
The first move is to strip friction out of your own checkout ruthlessly. The leading cause of abandonment, unexpected shipping, tax, and fees, improves dramatically simply by showing the full total early, allowing guest purchase, and compressing input steps. Shoppers who delegated discovery to AI arrive with clear intent, which may in fact make them less tolerant of friction along the way. Losing high-intent, pre-vetted traffic to a cumbersome purchase flow is the failure to avoid above all.
The second move is to make your own checkout agent-ready. As protocols like ChatGPT's Instant Checkout and AP2 spread, whether an agent can correctly read your product data, inventory, pricing, and payment options will determine whether you capture the opportunity. Even so, prioritize a design that keeps the purchase experience and customer data on your side. When a transaction closes entirely inside a third-party AI, you risk losing the very touchpoint you would use to build the next relationship. As Epsilon notes, that single interaction may be the only one you get, so returning value at every touchpoint matters.
The same picture holds in travel, transport, and lodging booking. AI takes over the advice and comparison of "where should I stay" and "which is the cheapest flight," and that layer will keep automating. But the core of the transaction, confirming the booking, taking payment, and handling later changes and issues, is worth keeping firmly in the operator's hands. Open discovery to AI while polishing the purchase experience according to your own design philosophy. That two-tier approach is the realistic play for the AI commerce era.
Summary
What The Current's analysis drives home is that framing the question as "will AI steal discovery from us" may miss the point entirely. For the foreseeable future, consumers use AI to research and press the buy button themselves. As long as cart abandonment sits above 70% and the upside from better checkout remains enormous, the decisive line is drawn at conversion, not discovery.
The next things to watch are how long AI-referred traffic keeps converting at elevated rates, and when agentic payments like ChatGPT's Instant Checkout and AP2 shift from "assisting" to "acting on behalf of" the shopper. Before that inflection arrives, cut the friction in your own checkout and build a design that does not surrender the customer relationship to third-party AI. That is the surest move available now to keep winning even after conceding part of the discovery layer.
