Google's I/O 2026 included an announcement that has not gotten the attention it deserves from performance marketers. Universal Cart — Google's AI shopping hub spanning Search, Gemini, YouTube, and Gmail — went live in the US last Tuesday. Paired with the Agent Payments Protocol, which now supports "Human Not Present" autonomous purchases, it suggests a near-future in which a customer journey that used to be ad-to-page-to-checkout becomes ad-to-feed-to-agent-to-cart. The landing page, the artifact most performance teams have spent a decade optimizing, may not be in the path at all.
Broader context: the bottleneck moved, not the problem
The post-click experience has been the quiet bottleneck of paid advertising for a decade. Platforms competed on creative production, audience targeting, and bidding sophistication. The page the user landed on was the marketer's problem, optimized inconsistently, governed rarely, and treated as a finishing step rather than a system. Universal Cart and AP2 do not change the bottleneck; they change who holds the bottleneck.
When a user adds a product to Universal Cart from a YouTube ad, Gmail summary, or Gemini chat, the cart object — not the landing page — becomes the next step. The cart pulls product data from the Shopping Graph, not from the page. It enriches with price history, deal alerts, and compatibility checks based on structured feeds. The page, when it appears at all, becomes a backup, a confirmation, or a place to read reviews. The decision happens elsewhere.
Meanwhile, AP2 v0.2.0's Human Not Present capability lets agents purchase autonomously within pre-set limits. The user no longer needs to be present for the conversion. They authorize a directive — "buy this when it drops below $80, ships in three days, compatible with my setup" — and the agent executes. The landing page is irrelevant to that transaction.
This connects to a broader pattern visible across the platforms. TikTok GO lets users book full holidays inside the platform. Meta's Advantage+ Leads Campaigns let advertisers engage prospects on Messenger without ever sending them to a page. Each of these is a small step away from the page as the locus of the post-click experience. Universal Cart is the most ambitious of them, but it is part of a wider movement, not an outlier.
Cautious prediction: three things likely in 6–12 months
Over the next 6 to 12 months, three things are likely.
First, the most successful brands on Google Shopping will probably treat product feed quality and ad creative congruence as a single discipline rather than two separate ones. The structured data that Universal Cart consumes — price, availability, shipping promises, attributes — needs to say the same thing the ad creative says. If the ad shows a free-shipping badge and the feed flags shipping costs, the cart surfaces the cost, and the alignment drift happens inside Google's UI rather than on the marketer's site.
Second, AI agents will likely become a measurable share of conversion events. Modest in the next 12 months — Gemini Spark's early beta rollouts suggest limited reach — but the share will be non-zero, and it will require attribution that distinguishes human-completed from agent-completed transactions. Teams that do not separate these in their dashboards will misread campaign performance.
Third, the governance question moves up the stack. Today, ad-to-page alignment is checked at the page level. In the agent-mediated future, alignment likely has to be checked at the protocol level — the ad, the product feed, the schema, the disclaimer text, and the structured promise that an agent will parse. The teams that build creative governance around this stack will compound their advantage; the teams that do not will accumulate invisible alignment debt across thousands of SKUs and campaigns.
Fourth, and most uncomfortable for measurement teams: attribution gets harder before it gets easier. A purchase initiated by an agent acting on a directive set three weeks ago, executed at a price point seven dollars below the one the ad showed, fulfilled through a checkout the customer never saw, is not the same event as a click-through purchase. Pretending it is the same event for the sake of clean dashboards will produce confidently wrong campaign optimization decisions. The teams that can taxonomize the conversion path — human-completed, agent-assisted, agent-completed — will spend the rest of 2026 making better decisions than the teams that lump them together.
The mistake to avoid: assuming this changes nothing because the page is still in the URL. The page may still load. It just may not matter for the conversion.
Philosophical and practical: the examined campaign
There is a useful, slightly old idea worth recalling here: clarity compounds. The platforms have made creative production faster than at any point in advertising's history. They have made audience selection faster, bidding faster, optimization faster. And now they are making the click-to-conversion path faster, by removing the page and the human from large parts of it.
What is left for performance teams to compound on is not the speed — that is commoditized. It is the alignment. Whether the ad's promise, the feed's metadata, the page's headline (when it appears), the disclaimer's required text, and the cart's surfaced price all say the same thing. What ad-to-page congruence actually means was already a useful concept at the page level; it gets sharper as the surfaces multiply.
AdAlign's product was built on the page-level version of this question: does your ad and your landing page tell the same story? Universal Cart and AP2 make that question larger, not smaller. The story now needs to hold across more surfaces, told by more systems, parsed by more machines. The governance gap does not close because agents are involved; it widens, because there are more surfaces to misalign across.
A concrete version of what this looks like in practice: imagine an ad creative shows a $79 product with a free-shipping badge. The landing page hero matches it. The product feed, updated by a different team on a different schedule, still lists $84 from last month's pricing test, and the shipping field encodes a $6 charge that the merchandising team forgot to clear after a holiday promotion. The user never sees the discrepancy on the page, because the page is honest. Universal Cart, parsing the feed, surfaces $84 plus $6 shipping. The user adds it to cart, sees the cost, and bounces. The ad creative team blames Google. Google blames the feed. The feed team did not know the ad creative existed. This is alignment drift at protocol level, and it is invisible to every team in the chain unless someone is checking the full surface stack.
The most concrete current example of this is covered in the companion piece: how to fix AI Max final URL expansion message-match drift in 48 hours, where algorithmic URL selection has already pulled the page out of the marketer's direct control.
The examined ad campaign — the one where every surface has been checked, where every promise is consistent across creative, page, feed, and disclaimer — is the one that will likely perform in this environment. It is also, increasingly, the one that is impossible to maintain by hand.
Where to start
If your team runs paid acquisition at scale, this is the moment to put alignment monitoring in place, before the agent-mediated surfaces compound the problem. Start with a free audit to see what your current alignment looks like. Teams running continuous campaigns should look at the Growth tier; agencies managing multi-client portfolios should look at the Agency tier and the agencies overview.