A retailer whose product is saved to Google's Universal Cart receives no request, no event and no count until the shopper presses transfer, according to an analysis published today by Brandon Lazovic, an e-commerce SEO consultant and data engineer. Working from Google's October 15 launch post, its May 19 preview and eight Universal Commerce Protocol and Merchant API guides, Lazovic found that Google holds the saved items itself, sends the merchant a single cart request at transfer, supports no updates or synchronization in this release, and starts its merchant reporting at "Clicks to Buy." A first-hand check of eight named launch merchants' public commerce profiles returned a file for two Shopify stores and a 404 for the other six.

In Short

When a shopper puts a product in Google's new shopping basket, the shop that sells it is not told anything until the shopper sends the basket over to the shop's own site. That matters to retailers because the moment they normally use for abandoned-cart emails and for counting interest now stays inside Google. What changes is that saves which never turn into a transfer will not appear in any report the retailer can open, so the shop sees only the shoppers who come through at the end.

A cart the retailer cannot see

Google launched Universal Cart in the United States on October 15, 2026, across Search and the Gemini app. PPC Land's report that day covered what the consumer post said and what it left out: YouTube and Gmail, which Google named in May, were missing, and the post gave no figures on fees, ranking or data sharing. Lazovic's analysis, dated October 16, goes a layer down, into the developer documentation that tells merchants what actually reaches their servers.

The short answer is: almost nothing, until the end. According to Lazovic, Google's Cart API reference states that "Google accumulates those items locally" and "does not fire multiple API calls as items are added." The merchant's cart endpoint, the server address that receives Google's request, gets one request carrying every saved item at once, and only "when the user clicks the transfer button."

That design has a direct commercial consequence, one Lazovic puts in plain terms. "Until then, the save-to-cart moment your abandoned-cart emails depend on stays with Google," he writes. A product added to the cart on a Monday and transferred on a Friday reaches the retailer, in his description, "as a Friday visitor who arrived with a full cart." A product saved and never transferred "appears in no report you can open."

Lazovic is explicit that this is read from the documentation rather than observed. In an accuracy note at the end of the post, he states that he has "not used Universal Cart as a shopper or run a transfer against a live store, so the reporting gap is read from the documentation and not observed in an analytics account."

One-way, by design

The request itself is narrow. According to Lazovic, Google's Cart API overview says "This initial release is limited to cart creation and transfer," and that "updating, deleting, or synchronizing cart contents between Google and your system are not supported." The merchant's system answers the single request with a link, which Google calls the continue_url, pointing to a cart on the merchant's own site already filled with the transferred items.

So the data flows in one direction. Google can send a cart; the retailer cannot send changes back, cannot see edits the shopper makes on Google's side, and cannot reconcile its own cart with Google's.

One Google page points the other way. Merchant Center's test environment, according to Lazovic, lists optional tests for reading, updating and canceling a cart alongside the test for creating one. His interpretation, labelled as his own read, is that "the test tooling is ahead of the shipped feature, because the overview rules out updates in this release." Google has not said whether, or when, update and sync support will arrive.

Where the reporting starts

Merchant Center does offer reporting for Universal Cart traffic, but it begins at the transfer rather than at the save. According to Lazovic, the Merchant Center dashboard for the Universal Commerce Protocol (UCP) lists seven metrics, the first of which is "Clicks to Buy." The Merchant API's alpha report for the same data adds cart_transfers, average cart value and average items per cart. "Neither report counts a product saved to the cart," Lazovic writes.

That leaves retailers with a familiar measurement problem in a new place. Saves are the upper-funnel signal - the equivalent of an add-to-cart on a merchant's own site - and they sit with Google. What the merchant sees is the conversion-adjacent end of the journey. Whether Google will add a save count to Merchant Center, or share aggregated save data in another form, is not addressed in any of the documents Lazovic cites.

PPC Land's October 15 report had already noted that the consumer post said nothing about "whether abandoned-cart signals reach the retailer." The developer guides, as Lazovic reads them, answer the question: in this release, they do not.

Tagging the transfer

Because the merchant's server writes the continue_url, the merchant controls what parameters it carries. Lazovic's recommendation, listed in his post as one of five pre-holiday tasks, is to add a parameter that marks the session as a Universal Cart transfer. Without one, he writes, "a shopper arriving from Google with a full cart lands in your reports as ordinary Google or direct traffic, the same blind spot ChatGPT's untagged citation links created."

The comparison has history. In June 2025, PPC Land reported that ChatGPT added utm_source=chatgpt to links in its "More" section, which had previously carried no tracking parameters and were often counted as direct traffic. Measuring AI referral traffic has been a recurring difficulty for analytics teams since, and Universal Cart adds a case where the referral arrives with a pre-built basket attached. Lazovic adds that a transfer which "arrives untagged cannot be told apart afterward," so the parameter only helps from the day it is added.

Price alerts and promo codes

Universal Cart's alerts and offers draw on data Google already holds. According to Lazovic, the launch post says the Shopping Graph works "in the background, sending notifications about price drops or when an item comes back in stock." Google describes that index as "60+ billion continually refreshed product listings."

What Google has not said is how large a drop triggers an alert. Lazovic treats the threshold as unknown and flags one possibility as "Suspected, not confirmed": a sale price that has ended in a store but still sits in Google's copy of the merchant's data could trigger a price-drop alert, sending the shopper to a higher price than the notification implied. The concern is not hypothetical in kind. In September, PPC Land reported on a Merchant Center import that showed products nine to ten days out of date despite a stated 24-hour refresh.

Promo codes have a partly documented source. Google's guide to promo codes in UCP checkout, as quoted by Lazovic, names two kinds of promotion. "Auto-applied promotions" are "Offers ingested directly from your Google Merchant Center feeds," which Google "automatically applies." "User-applied promotions" are codes the shopper types in, and the checkout accepts up to 10 typed codes per session.

That guide, however, covers checkout sessions, which come after the cart. The launch post says the cart shows "relevant promo codes right from the cart" and does not say where those codes originate. Lazovic lists as unconfirmed "whether a code you never gave Merchant Center can appear there." The question matters beyond retailers. When the cart was first shown in May, PPC Land examined how automatic deal-finding could undercut coupon and deal sites that earn affiliate commissions on a last-click basis. Neither the launch post nor the developer guides, as Lazovic reports them, resolve where in-cart codes come from.

The Google Wallet card suggestion is a separate matter. Lazovic calls it "the one part of the cart a retailer has no input into at all."

Two checkout paths, two sets of requirements

According to Lazovic, Merchant Center's UCP integration hub gives a merchant two routes for Universal Cart, and they ask very different things of the merchant's feed and code. The hub is rolling out gradually in the US. PPC Land reported on its launch on October 6, when Google made cart transfer, native checkout and identity linking available through Merchant Center and pushed Canada and Australia to 2027.

Transfer to merchant cart sends the items to the retailer's own site. It requires one cart endpoint and matching product identifiers. According to Lazovic, Google says this path "doesn't require full product feed setup, product warnings, or customer support information," and describes it as the route that "Minimizes implementation effort to accelerate UCP integration." Checkout tracking stays on the merchant's own pages.

Native checkout on Google keeps the purchase on Google with Google Pay. It requires three checkout endpoints, order-status updates sent back to Google, and an accepted program addendum. Each product needs a native_commerce(checkout_eligibility) attribute set to true, and the merchant must provide a return policy and support contact in Merchant Center. The attribute, according to Lazovic, "defaults to FALSE," so a product never flagged stays out.

Several product types are excluded from native checkout altogether. Lazovic lists subscriptions, personalized goods, final-sale and pre-order items and age-restricted goods among them, citing Google's account preparation guide. For those items, cart transfer is the only path. Native checkout also hands rendering to Google: the checkout page is Google's, not the retailer's.

Google's native checkout overview adds a line on how that path behaves. "Initially, Google will render the user interface for the buyer, with future plans to support more agentic experiences," it says, according to Lazovic, and once payment starts, "The Agent is not involved in this part, ensuring determinism."

Eight merchants, two profiles

Lazovic also tested which paths some of the named launch merchants declare. A store's UCP profile is a public file at /.well-known/ucp that lists the capabilities the store supports. On October 16, he fetched that file for eight merchants Google named across its May and October posts.

Two returned a profile. Steve Madden and Fenty Beauty, both Shopify stores, publish the same list of 11 capabilities, including the cart capability, the checkout capability and a Google Pay payment handler, which means both declare support for both paths. Lazovic notes that two stores "cannot show what every Shopify store publishes, though the identical lists suggest Shopify generates the file." That would fit what PPC Land reported in September, when Google matched Shopify merchants to their Merchant Center accounts and switched native checkout on automatically, though Lazovic does not make that link himself.

The other six - Nike, Target, Ulta Beauty, Walmart, Wayfair and Quince - returned 404 at that path.

This result needs careful handling, and Lazovic handles it that way. It came from one fetch per merchant, made on October 16 "from a commercial virtual private network (VPN) exit," and he states that a 404 "means only that no file sits at that path." Merchant Center asks each merchant where its profile lives, so a 404 at the standard path, in his words, "says the file is elsewhere, absent or blocked from my connection, and nothing about whether the retailer is integrated." PPC Land has not repeated the fetches, and the result remains Lazovic's own unverified test.

A tension with the documented requirement

The finding still sits awkwardly beside Google's own instructions. Lazovic notes that Google's cart guide tells merchants to publish the profile at that path, "which makes the six misses worth a second look from a different connection." PPC Land's October 15 report, drawing on the October 6 Merchant Center hub, stated that cart transfer requires "a UCP profile at /.well-known/ucp" along with the cart capability and UCP version 2026-04-08 or later.

Two readings are possible, and the available material does not settle which applies. Either the large retailers host their profiles at a location registered with Merchant Center rather than the standard path, or block requests like Lazovic's, or several named launch partners had not published the file that Google's guide describes on the day after launch. Google has not commented on the discrepancy. A similar pattern appeared in May, when Originality.ai found 26 detectable UCP implementations among more than three million websites and none of the five co-developers had a public UCP profile on its main domain; that study also cautioned that large retailers might connect in ways a public scan would not detect.

Building to an older version

The two profiles Lazovic retrieved carry one more detail. Both list the August UCP release (2026-08-25) first and still support the April release (2026-04-08). According to Lazovic, the April release is the version Google's Cart API guide tells merchants to handle, while the August release "rewrote fulfillment, consent and signing fields."

The consequence, as he puts it: "A merchant building to Google's guide in October 2026 builds to the older shape." The protocol's governance has widened over the same period; on April 24, Amazon, Meta, Microsoft, Salesforce and Stripe joined the UCP Tech Council alongside Google, Shopify, Etsy, Target and Wayfair. A standard moving on a quarterly cadence, with Google's own merchant guide one release behind, adds a maintenance cost that the launch post does not mention.

On performance, Google's guide recommends that the cart endpoint answer "ideally in under one second" and return a 429 (too many requests) response when traffic exceeds the merchant's limits, according to Lazovic. Holiday peaks will test that.

Agentic in May, manual in October

Lazovic also compared the May 19 preview with the October 15 launch post line by line. The comparison overlaps with PPC Land's October 15 report, with small differences in wording. Lazovic says the launch came "about a season later than promised"; PPC Land described it as nearly five months after the I/O preview, after the summer window had closed. Both describe the same timeline.

His table lists four changes. The preview covered Search, Gemini, YouTube and Gmail, with "YouTube and Gmail to follow"; the launch covers Search and the Gemini app. The preview promised "insights on price history"; the launch post lists price-drop and restock alerts with no price history. The preview described flagging incompatible products, such as computer parts that will not fit together; the launch post does not mention it. The preview said Wallet knows "payment method perks, loyalty information and merchant offers"; the launch post names card perks and does not mention loyalty.

Lazovic places Universal Cart in AI Mode in Search, while Google's launch post, as reported by PPC Land, says the cart "is rolling out in the U.S. across Google Search." Google's own usage example has a product saved in AI Mode, so the difference is one of emphasis rather than conflict.

The framing changed more than the features. At I/O, according to Lazovic, Google introduced the cart under the line "We're building the foundation for agentic commerce" and called it "the next step that brings this all together." The October post splits Google's commerce work into "agentic capabilities that act on your behalf" and "practical tools that give shoppers confidence and control," and says Universal Cart "is built for that second track."

Lazovic cautions against reading too much into omissions. "Absence from a launch post does not prove a feature is missing," he writes. What the posts do confirm, in his reading, is that "the product that shipped is a cart a person fills by hand on Google's surface," which leaves retailers with the levers they already own: "its feed and its checkout integration."

Why this matters for marketers

The practical effect is a shift in where intent data accumulates. In late May, Google's Nick Fox described cart-stage merchandising as interesting for retailers but declined to confirm ad plans for Universal Cart. If saves stay with Google and only transfers reach the retailer, Google holds the most complete picture of which products shoppers are weighing across stores - exactly the data that would make cart-stage advertising valuable. Retailers, by contrast, see a filtered end of the funnel.

For paid search and shopping teams, two dependencies follow. Alert accuracy rests on feed freshness, because alerts are generated from Google's copy of the merchant's data. And attribution rests on whatever the merchant writes into the continue_url, because Google's documented reports do not count saves and an untagged transfer is indistinguishable from other Google traffic.

For the eight named merchants Lazovic tested, the open question is narrower and verifiable: whether the six 404s reflect profiles registered elsewhere, blocked requests, or files not yet published. Google's documentation points to the standard path. The answer would say a good deal about how far the launch partners' integrations extend beyond the press materials.

About the author of the analysis

Lazovic describes himself on his site as an "E-commerce SEO Consultant & Data Engineer" with "a decade of experience driving enterprise-level organic growth for Fortune 500 brands." His About page lists his current role as SEO Data Engineer at BrightEdge, a position held since 2025, with earlier roles at Pella, U.S. Bank, Rocket Mortgage, CDK Global and General Motors. The page states that where a finding is his inference rather than a source's claim, he labels it as such, and the Universal Cart post follows that practice, marking several points as "My read," "Suspected, not confirmed" or "Unconfirmed."

Timeline

Summary

Who: Brandon Lazovic, an e-commerce SEO consultant and data engineer who lists BrightEdge as his current employer, analyzing Google's Universal Cart and its developer documentation. Retailers named by Google, including Nike, Target, Walmart, Wayfair, Ulta Beauty, Quince, Steve Madden and Fenty Beauty, appear in his profile test.

What: According to Lazovic's reading of Google's guides, Universal Cart sends a merchant one cart request at transfer and none while items are saved; this release supports only cart creation and transfer, not updates or sync; Merchant Center reporting starts at "Clicks to Buy" and does not count saves; native checkout requires a per-product eligibility attribute that defaults to false; and the checkout accepts up to 10 typed promo codes per session. His single VPN fetch found UCP profiles for two of eight named merchants, a result in tension with Google's guidance that profiles sit at /.well-known/ucp.

When: The analysis and profile test date from today, October 16, 2026, one day after Google's October 15 US launch and five months after the May 19 I/O preview.

Where: The United States, across AI Mode in Search and the Gemini app, with merchant setup in Google Merchant Center.

Why: Saves that never reach the retailer move the upper part of the shopping funnel inside Google, leaving merchants to measure only transfers, and only if they tag the continue_url themselves.