When the price in a brand’s product feed doesn’t match the price on the page that feed points to, the product can be pulled from consideration before a single shopper or agent ever evaluates it. Google’s Merchant Center makes this explicit: its crawler compares the price attribute in the data feed against the price on the landing page, and when the two disagree, the product may be disapproved (Google, n.d.-a). Not deprioritized. Removed. The brand did nothing wrong on positioning, nothing wrong on differentiation, nothing wrong on trust. It lost the listing to a synchronization gap.
This is the dimension the Brand Visibility for Agentic Commerce (BVAC) framework calls Latency and Data Freshness, and it’s the one most likely to be doing real damage in a brand that has done everything else right. It isn’t about whether the data exists — that’s Attribute Completeness, a prerequisite handled elsewhere. The product here has a price and an availability status. They’re just stale, or they disagree across surfaces, and that’s enough.
The disqualifiers are operational, not strategic
The failure modes read like a sysadmin’s checklist, which is the point. Price is the most common one: Google preemptively disapproves products when the feed price and the landing-page price don’t match, erring toward removal on anything that looks like a violation (Google, n.d.-b). Availability is the second, and the consistency bar is higher than most brands realize — the status has to match across the landing page, the checkout page, the structured data markup, and the feed, and a mismatch in any one of them triggers a disapproval (Google, n.d.-c).
The third is the one that catches engineering teams off guard. Google’s crawler reads the HTML returned by the web server. If the price is written into the page by JavaScript after the page loads, the crawler doesn’t see it, and that alone trips a mismatch error (Google, n.d.-a). A brand can have a perfectly correct price that a machine reader simply never receives, because the number arrives a few hundred milliseconds too late and on the wrong side of the render. The fix is to put the price in the server-rendered HTML, which is plumbing, not merchandising.
This is the agentic surface, not the legacy one
It’s tempting to file all of this under old Shopping-ads housekeeping and move on. That would be a mistake, because the Merchant Center feed is the same product-data layer Google’s agentic surfaces draw on — Universal Commerce Protocol (UCP) onboards retailers through Merchant Center, and the buy experience in AI Mode and Gemini runs on that feed. The consistency rules that govern whether a product appears in a Google Shopping listing are, increasingly, the rules that govern whether it reaches an agent acting on Google’s surfaces. A product disapproved for a price mismatch isn’t in the set the agent chooses from, and the agent never knows it was supposed to be.
More broadly, the agentic protocols are converging on exactly this surface. ACP standardizes product feeds for AI agents and UCP defines a machine-readable capability and catalog layer, which means the data an agent reads is moving from a rendered webpage it has to interpret toward a structured feed it consumes directly. That doesn’t relax the freshness requirement. It tightens it, because a structured feed is read literally — there’s no human to notice the price looks off and check the cart.
Freshness is a trust signal in machine-time
Reframe the dimension and it stops looking like maintenance. To a human shopper, a price that’s thirty minutes stale is a minor annoyance, caught and corrected at the cart. To an agent assembling a shortlist against a latency budget, an inconsistency between what the feed says and what the page shows is a reliability defect, and the system’s response is to withhold the listing rather than surface it and sort the discrepancy out later. The Merchant Center crawler’s behavior is the documented version of an instinct agents share: when the signals conflict, don’t recommend.
Latency is part of the same story. The frontier-agent study that mapped how shopping models choose noted that its authors had to cap the models’ reasoning depth precisely because latency and cost are binding constraints in real-time commerce (Allouah et al., 2025). An agent operating under a time budget isn’t going to wait out a slow-responding feed or re-fetch a page that hung. A surface that’s slow or stale is, for that decision, absent — and the brand sees nothing, because a product that was never in the set generates no signal it lost. It’s decision invisibility again, with a cause that lives in the propagation pipeline rather than the catalog.
Listings aren’t static assets
There’s a constructive corollary. The same study showed the agent-facing surface is responsive: a single well-aimed edit to a product’s description moved its selection share for many of the models tested (Allouah et al., 2025). But responsiveness depends on the current state actually reaching the surface the agent reads. An optimized listing the agent never sees — because the feed disagreed with the page and the product was pulled — returns nothing. Freshness isn’t the glamorous half of the work, but it’s the half that determines whether the glamorous half is visible at all.
Where to start
Picture a brand that scores well across the board — identifiers resolve cleanly, the category-standard attributes are present, the review signal clears the floor — and that runs promotions through a feed syncing on a schedule that lags the live site by tens of minutes. On an ordinary day this never surfaces. On the day that matters, when a sale goes live and agent traffic spikes with it, the feed shows yesterday’s price while the page shows the markdown, the crawler flags the mismatch, and the products are disapproved during the exact window the brand was counting on. Every other dimension scored well and the composite collapses anyway, because the framework sets the composite at the weakest effective dimension and this is now it.
The remediation isn’t a strategy deck. It’s a feed-sync cadence fast enough to keep the feed and the page in agreement during price changes, price and availability written into server-rendered HTML rather than injected by client-side script, and an inventory or content API for the categories that change prices intraday. None of it is interesting work. All of it determines whether the interesting work is ever seen.
The dimension that decides visibility for everything else is the least glamorous one in the framework. A brand can win every argument it has about positioning and differentiation and still lose, quietly and during peak demand, on a data-synchronization job that no one in the room thought was their responsibility.
References
Allouah, A., Besbes, O., Figueroa, J. D., Kanoria, Y., & Kumar, A. (2025). What is your AI agent buying? Evaluation, biases, model dependence, and emerging implications for agentic e-commerce (arXiv:2508.02630). arXiv. https://arxiv.org/abs/2508.02630
Google. (n.d.-a). How to fix: Mismatched product price. Google Merchant Center Help. https://support.google.com/merchants/answer/12159029
Google. (n.d.-b). Issues in Merchant Center. Google Merchant Center Help. https://support.google.com/merchants/answer/12153802
Google. (n.d.-c). How to fix: Inaccurate availability status due to inconsistency between feed and landing page. Google Merchant Center Help. https://support.google.com/merchants/answer/9773127
