For most of the past decade, e-commerce brands spent heavily to win customers before they reached checkout. Search ads, social campaigns, influencer partnerships, and email sequences. The logic was sound: control the discovery journey, and conversions will follow. That logic is now fraying.
AI tools are steadily compressing the discovery phase. Shoppers who once clicked through five or six product pages and comparison tabs now describe what they need inside ChatGPT or Perplexity and receive a curated shortlist in seconds. According to data cited by Similarweb, 38% of U.S. internet users now use AI tools for product discovery, up from just 8% two years ago. EMARKETER projects that AI-platform-driven e-commerce sales will reach $20.9 billion in 2026 alone, nearly four times the prior-year figure.
As that upstream terrain shifts, one question is quietly reshaping how serious e-commerce operators think about their entire tech stack: if the discovery journey is compressing into AI tools that brands don't control, what surfaces do they actually own?
The answer, increasingly, is checkout.
Checkout Is Inheriting the Entire Journey
Rokt, the commerce technology company that powers transaction experiences for brands including PayPal, Uber, Live Nation, and Wayfair, has been making this argument for years. Its recently published white paper, The New Economics of Checkout, offers the clearest articulation yet of what that shift means in practice.
The paper frames the period spanning selection, cart or review, payment, and confirmation as the "Transaction Moment™"; the four pages where customer attention, trust, and intent converge at their peak. The argument is that as AI tools absorb more of the upstream discovery work (search, browse, comparison, research), the Transaction Moment stops being just a place to process payments. It becomes the only moment in the journey where a brand still controls the experience directly.
That weight was already significant. Now it's heavier.
"The storytelling, the basket-building, the cross-sell seeding, the relationship touchpoints," the white paper states, describing the work that upstream pages once carried. "All of it now has to land on the pages where the transaction happens."
The numbers behind checkout mismanagement make this shift feel urgent. Baymard Institute's research, drawn from aggregate analysis of 50-plus studies, puts the global cart abandonment rate at 70.19%. That translates to roughly $260 billion in recoverable lost orders in the U.S. and EU markets alone. Baymard's separate checkout UX work found that the average large-scale e-commerce site could increase its conversion rate by 35.26% simply by improving checkout design. The revenue sitting uncaptured on these pages is not marginal.
The Instinct That Backfires
When brands first register the shrinkage of upstream surfaces, the instinct is predictable. They load more offers onto the cart page. They turn the payment screen into a cross-sell billboard. They pack the confirmation page with promotions. More placements, more revenue, the math seems straightforward.
The Rokt white paper documents why this instinct reliably backfires. The risk isn't just cart abandonment, though that risk is real. The deeper problem is emotional. Harris Poll research conducted on behalf of Rokt across 7,061 adults in six markets found that 73% of consumers report genuine joy in checking out online. Fifty-two percent said seeing their purchase confirmed is the single happiest moment of the entire transaction process. Eighty-nine percent found delight in completing long-desired purchases.
That emotional peak is what brands are putting at risk when they treat checkout as a promotional slot. The same research found that 74% of consumers would rather receive no offer at all than an irrelevant one. Sixty-two percent said they would abandon their cart in response to irrelevant content. And 41% said a delightful checkout experience would bring them back to shop again.
The paradox is sharp: the moment shoppers are most receptive, most trusting, and most likely to act is also the moment most vulnerable to being ruined by clutter.
Classic choice architecture research reinforces the point. Iyengar and Lepper's well-documented study found that consumers shown six options were ten times more likely to purchase than those shown twenty-four. Complexity at the point of commitment doesn't communicate abundance. It produces paralysis.
Four Pages, Four Different Businesses
One of the most useful frameworks in the Rokt white paper is its treatment of the Transaction Moment not as a single surface but as four distinct customer mindsets in sequence.
On the selection page, the customer is forming a basket. But AI compression is narrowing this phase; shoppers increasingly arrive having already decided on a product, which means the selection page is shifting from broad discovery toward validation. The bar for relevance is higher because the customer already has a frame of reference. Irrelevant offers that interrupt decision flow or introduce competing options can unravel decisions already close to being made.
The cart or review page is a confidence-building moment. The customer is validating choices already made. What works here are offers that complete the decision rather than reopen it. Highly relevant, low-friction add-ons that strengthen the chosen basket. What destroys value are basket-competing choices or a cluttered menu that re-triggers doubt.
Payment is the highest-stakes page in the entire sequence. The customer is committing funds. The white paper notes that payment and selection pages are suppressed at a rate roughly seven percentage points higher than confirmation or upsell pages in Rokt's network data, reflecting just how unforgiving this step is. Even minor friction punishes more here than anywhere else.
Confirmation, by contrast, is where satisfaction peaks. The commitment is complete. This is often the first meaningful moment of direct contact between brand and customer, and it's where loyalty enrollment, app downloads, and post-purchase engagement actions can genuinely start a relationship rather than interrupt one.
The takeaway is that treating checkout as a single uniform "checkout surface" is itself a strategic error. Each page carries different economics, different customer psychology, and a different risk profile for the same offer.
Suppression as a Revenue Strategy
The Rokt white paper introduces a concept that runs counter to how most e-commerce monetization is measured: suppression. When the system cannot identify a relevant offer for a specific customer on a specific page, the best decision is to show nothing at all.
This is not intuitive for organizations accustomed to measuring monetization by impression volume. But the behavioral logic is compelling. A customer who encounters an irrelevant offer during a high-stakes moment doesn't just decline it. They return with a slightly lower willingness to engage. Repeated low-quality exposures compound. The engagement asset deteriorates.
A customer who encounters no offer on a visit where nothing was relevant enough returns with their willingness to engage intact. Suppression, in this framing, is an investment in the long-term performance of the surface rather than foregone revenue.
Rokt's own network data shows the pattern at scale. Across eligible placements analyzed in the white paper, 28% were suppressed network-wide in a single-week snapshot. On payment and selection pages, suppression rates ran approximately seven percentage points higher than on confirmation pages, reflecting the higher sensitivity of those steps.
The Commerce Media Angle
The Rokt white paper situates checkout monetization within a broader argument about where value is concentrating in digital commerce. On-site sponsored search, historically the highest-margin segment of retail media, depends entirely on human browsing behavior. When AI agents research, compare, and select on behalf of consumers, they bypass sponsored listings entirely. That's not a future scenario; it's already happening, and the retail media landscape is already absorbing its effects.
Retail media networks proliferated between 2020 and 2022. Most failed to build durable adoption. The platforms that lasted were those with a genuine performance signal. The same dynamic is playing out now with commerce media more broadly: the surfaces that hold up are the ones where brands still control the experience, and customers still make purchase decisions with human intent.
Reuters reported in January 2025 that Rokt completed a $335 million secondary share offering with Tiger Global Management at a $3.5 billion valuation. Sacra's research estimates that Rokt hit approximately $900 million in annualized revenue by October 2025, up from $600 million at the end of 2024. The company serves more than 33,000 active clients, including more than half of the 200 largest e-commerce companies globally, and its Rokt Brain AI engine analyzes 1.95 trillion data points annually to deliver real-time relevance across the Transaction Moment.
Bloomberg coverage from January 2025 described the secondary sale. The growth numbers attached to that announcement, 43% year-over-year revenue growth reaching $600 million, reflect an enterprise built around a thesis that has only grown more relevant: checkout is a media surface, and the brands that govern it deliberately will capture value that more passive operators will miss.
According to Rokt's own data, the company is projected to power more than 10 billion transactions in 2026, across 165 million monthly active users globally. That scale creates a network effect: every transaction processed through Rokt's platform improves the model's predictions for all participants on the network, a compounding advantage that isolated checkout tools cannot replicate.
What the Next 18 Months Look Like
The Rokt white paper is deliberate about the timeline. AI-mediated commerce currently accounts for roughly 1.5% of total retail e-commerce in 2026, according to EMARKETER's estimates. Full agentic checkout, where AI agents complete purchases autonomously, remains farther out than early projections suggested. OpenAI scaled back its plans to enable in-chat purchases in March 2026 after finding that users researched products inside ChatGPT but didn't complete purchases there, redirecting shoppers instead to retailer sites and apps.
That pullback, as Forrester noted in response, actually reinforces the strategic logic. Discovery is compressing into AI. The transaction itself is staying where brands control the experience. The checkout pages are inheriting more responsibility, not less.
The Rokt white paper identifies utility purchases, household goods, personal care, and groceries as the categories where AI-mediated discovery is gaining traction fastest. These are the products consumers buy repeatedly with well-defined preferences. For e-commerce businesses operating in these categories, relevance infrastructure at the Transaction Moment is not a future investment. It's a current operational need.
Considered purchases, where browsing is part of the value itself, will compress more slowly. But the paper argues that most organizations are underestimating how soon the shift will arrive, even there.
Four Practical Moves
The Rokt white paper closes with four specific recommendations for e-commerce leaders evaluating their Transaction Moment strategy.
The first is to audit each page against the upstream work it now absorbs, not just its original function. Selection, cart, payment, and confirmation carry different economics. Treating them as a single checkout surface produces generic approaches that underperform on every page.
The second is to establish showing nothing as a first-class outcome. If the system cannot identify a relevant action for a specific customer on a specific page, the best decision is no action. This requires configuring decisioning systems with a minimum relevance threshold and measuring suppression rate as a health indicator rather than treating it as a gap.
The third is to set relevance guardrails before scaling. Frequency caps, suppression rules, and brand controls are protections for the engagement equity that make the transaction moment commercially valuable. Organizations that define these after scaling typically find the damage already done.
The fourth is to measure longitudinal outcomes, not just clicks. Click-through rate captures a single interaction. The value of these pages compounds or deteriorates over time. Repeat purchase intent, engagement decay over time, conversion protection, and the revenue impact of suppression are the metrics that reveal whether a strategy is building an asset or depleting one.
A Surface That Was Always There
The economic case for treating checkout as a strategic surface is not new. What has changed is the urgency. As AI tools absorb more of the discovery journey, checkout pages are no longer just the last step in a longer journey. They're the longest conversations most brands will have with their own customers.
E-commerce Fastlane's coverage of Rokt's platform puts the revenue stakes in direct terms: Rokt's confirmation page product generates up to $500,000 in incremental profit per one million transactions, while the payment page product generates up to $400,000 per one million transactions. These are not projections for edge cases. They are performance benchmarks from a platform operating at network scale.
The brands building relevance infrastructure at the Transaction Moment now are not just responding to a trend. They're accumulating years of machine-learned insight about what works for which customer, in which context, at which point in the journey. That advantage is not available to organizations that decide to start later.
