Skip to main content
Available on Shopify The Shopify app that audits and fixes what AI reads about your products. On Shopify Install free
Shopify

Shopify sales: Google and AI agents see them, your product pages don't

10 min read Recently updated
#shopify #structured-data #strikethrough-price #shopify-theme #study
Share

When a Shopify store runs a sale, the merchant enters a compare-at price in the admin and the product page shows it. But do Google, AI agents and tools that read the page receive that information? On Shopify, the answer depends on four layers that do not say the same thing. We measured them on October 13, 2026 across 1,125 product pages from 379 French Shopify stores.

The question is getting urgent: since September 28, 2026, Google has been rolling out monitoring in AI Mode to all users, which can alert someone to a price drop or a restock by continuously checking the web and its Shopping Graph, according to Robby Stein, Google’s Vice President of Product for Search. Google has not said whether it reads product page markup or the Merchant Center feed. The four layers below are the ones a Shopify sale goes through.

What we measured

The sample follows the method of our Shopify GEO barometer, with a list rebuilt on the day from the clients of the 133 French Shopify partner agencies in the official directory. 379 Shopify stores could be analyzed. For each, three products drawn at random, reproducibly, for 1,125 product pages.

We compared the real variants, read from the public JSON Shopify serves for every store, with each page’s Offer markup. We also recorded the Shopify theme of 337 stores (Shopify writes it into every page), rendered the 132 product pages on sale in a browser, JavaScript included, queried the UCP channel for AI agents for each of them, and compared with the WooCommerce sites from the same lists. Our requests carried our bot’s identifier and respected each store’s robots.txt. No store is named.

Layer 1: the Shopify compare-at price has no end date

In Shopify, a sale is expressed with the compare-at price, set above the selling price. That is what our study reads: 132 of the product pages with markup had at least one variant in that state.

Shopify’s public product JSON has no field for a sale end date. To schedule a start and an end, apps (including Shopify’s own Launchpad, for Shopify Plus) change the prices on the chosen dates. Shopify discounts (codes, automatic discounts) do have an end date, but they apply at cart, not to the product page price. In the product data, a Shopify sale therefore has no end date: only a compare-at price, present or absent.

Layer 2: the Google feed and the AI agent channel carry the sale

According to Google’s help page on Shopify, when the compare-at price is higher than the price, the Google & YouTube app sends the sale to Merchant Center, provided automatic product sync is turned on. Prices update on Google within 24 hours.

The Merchant Center feed can also date a sale with the sale_price_effective_date attribute. Since Shopify stores no end date, our reading is that the feed sent from Shopify has none, unless a feed tool adds it. The Google documentation we checked does not say.

For a store using the Google & YouTube app with automatic sync on, Google therefore receives the sale price through the feed.

Shopify opens a second machine channel: the Universal Commerce Protocol (UCP), announced by Shopify and Google on January 11, 2026 so that AI agents can browse the catalog and place orders (see our article on AI Mode Shopping and Gemini for Shopify). Shopify stores expose it at /api/ucp/mcp. We queried it on October 13 for the 132 product pages on sale, with the get_product tool and the example agent profile from Shopify’s documentation:

UCP channel, products on saleResult
Responses received129 of 132
Euro responses carrying the strikethrough price (list_price)107 of 107
Of which amount identical to the Shopify compare-at price101 of 107
Responses with a sale end date0 of 129

The other 22 responses were in a currency other than the euro and were not compared; the 6 gaps concern the top amount of the range, not the presence of the strikethrough price. The agent channel therefore knows the strikethrough price, not when the sale ends. At Google, product discovery goes through the Merchant Center feed, according to its UCP documentation: the store’s UCP address serves agents that query it directly.

The page, for its part, does not declare the sale.

Layer 3: the product page describes stock well, never the sale

The Offer markup on Shopify product pages describes stock and price correctly:

CheckResult
Product pages with Offer markup998 of 1,125, or 88.7%
Multi-variant pages that describe every variant406 of 429, or 94.6%
Partially sold-out pages flagging sold-out variants101 of 107, or 94.4%
Fully sold-out pages still marked “in stock”2 of 141, or 1.4%
Marked-up price matching no real variant11 of 840, or 1.3%

The sale, however, disappears. Google documents how to declare it on the page: in merchant listing structured data, the active price becomes a sale price when the Offer includes the original price, marked with priceType https://schema.org/StrikethroughPrice (Google still accepts ListPrice for a transition period).

None of the 132 product pages on sale did this. Not in the page source, not after JavaScript ran in a browser, not in microdata. The page shows the strikethrough price to the visitor and leaves it out of the structured data. It is technically possible, though: in one store, an SEO app adds a ListPrice to the markup, but on products that are not on sale.

ThemeStoresPages on saleStrikethrough price in markup
Dawn (Shopify)46220
Prestige2690
Impact16120
Horizon (Shopify)1050
Impulse1070
Focal950
116 other themes220720

This zero has more than one origin. On 59 of the 132 product pages on sale, the product markup is generated by Shopify itself: Dawn and Horizon, like other themes, write it with the structured_data Liquid filter. According to its documentation, the offer it outputs contains availability, price, currency and URL, with no strikethrough price and no date. For these stores, the gap comes from the platform, not the theme. On the other 73 pages, the markup is written by the theme or an app. 13 pages carry both, meaning two descriptions of the same product.

Layer 4: priceValidUntil is a theme formula

Google defines priceValidUntil as the date after which the price will no longer be available, and warns that a listing “may not display” if that date has passed. On Shopify, the date does not come from the sale: it is recalculated on every page view, depending on the theme or the app.

Formula observedWhere
Today + 10 daysPrestige (8 of 26 stores), Impact (8 of 16), Impulse (9 of 10), Focal (7 of 9), Expanse (4 of 8), Warehouse (3 of 3)
TomorrowBroadcast (6 stores out of 6), Pipeline (2 of 5)
Today + 30 days, + 90 days or + 1 yearOther themes (Empire and Ella: + 1 year) and SEO apps (8 of the 11 stores at + 90 days)
December 317 stores, various themes including Dawn
AbsentHorizon, Symmetry, Enterprise, Debut, and 37 of 46 Dawn stores

These dates describe no offer. Only one store in the study showed a date that had already passed, by 46 to 115 days depending on the page.

What about outside Shopify?

Declaring the strikethrough price is not out of reach for an ecommerce platform. WooCommerce’s code has done it natively since version 9.5.0, released on December 16, 2024: for a simple product on sale, the offer carries the regular price as ListPrice next to the sale price, and its priceValidUntil uses the sale end date when the merchant has set one, a native WooCommerce field. For a product with variations, it outputs no strikethrough price, only a sale price marked SalePrice, a value Google’s documentation does not support.

We checked it the same day on the 14 WooCommerce sites found among the sites from both agency lists that are not usable Shopify stores, through their public API of products on sale (Store API):

WooCommerce product pages on sale with Offer markupStrikethrough price in markup
Simple products7 of 10
Products with variations0 of 8

All carried WooCommerce’s default priceValidUntil, December 31 of the following year: none declared a sale end date. The field exists, it still has to be filled in.

Our reading: on Shopify, the gap is not a technical impossibility but comes from the data model (no end date) and from the markup produced by the platform or the theme.

What it changes

Where a Shopify sale lives on October 13, 2026: the admin knows the compare-at price but has no end date; the Google feed carries it when automatic sync is on; the UCP channel for AI agents returns it in 107 of 107 euro responses, with no end date; the product page declares it on none of the 132 pages on sale.
Figure 1: the four places of a Shopify sale. The machine channels carry the strikethrough price, the page does not, and none carries the sale end date.

Our reading: a Shopify store’s sale exists for machines that go through the platform (the Google feed, the UCP agent channel), but not for anything that reads the page. Shopify even contradicts itself: its UCP channel passes on the strikethrough price, its structured_data filter leaves it out of the page. Yet Google combines the page and the feed for its product rich results, and the page is the only source for tools that read the site directly. A price drop appears there as a new regular price, without the original price or an end date.

The fix, tested

For markup written in the theme, adding the strikethrough price to the Offer of each variant on sale is enough. Here is the offers block we wrote for this study, modeled on Google’s example, to place inside the markup’s Product:

"offers": [
  {%- for variant in product.variants -%}
  {
    "@type": "Offer",
    "sku": {{ variant.sku | json }},
    "price": {{ variant.price | divided_by: 100.0 | json }},
    "priceCurrency": {{ cart.currency.iso_code | json }},
    "availability": "https://schema.org/{% if variant.available %}InStock{% else %}OutOfStock{% endif %}",
    "url": {{ request.origin | append: variant.url | json }}
    {%- if variant.compare_at_price > variant.price -%},
    "priceSpecification": {
      "@type": "UnitPriceSpecification",
      "priceType": "https://schema.org/StrikethroughPrice",
      "price": {{ variant.compare_at_price | divided_by: 100.0 | json }},
      "priceCurrency": {{ cart.currency.iso_code | json }}
    }
    {%- endif %}
  }{% unless forloop.last %},{% endunless %}
  {%- endfor -%}
]

Two traps. Shopify gives prices in cents: divide by 100.0, not 100, because Liquid returns an integer when the divisor is an integer, and €39.99 would become 39. And on a theme that calls the structured_data filter, this block must go into markup that replaces the filter’s output, not into a second block.

We rendered it with the liquidjs Liquid engine on the real data of the 132 products on sale in the study: 561 variants, 469 of them on sale, all with a StrikethroughPrice equal to their compare-at price, and no variant off sale marked by mistake. liquidjs does not reproduce Shopify’s integer division: after deploying, check a product page with Google’s Rich Results Test.

What to check on your Shopify store

  1. In the Google & YouTube app, check that automatic product sync is on: it is what sends your compare-at prices to Merchant Center.
  2. On a product page on sale, open the page source (Ctrl+U, not the element inspector) and search for StrikethroughPrice. If it is missing, your page does not declare the sale, like every page in our study.
  3. Identify your theme (Online Store, then Themes) and compare it with the tables above. If the theme uses the structured_data filter (Dawn and Horizon do), the markup comes from Shopify: it has to be replaced with a block written in the theme, not supplemented with a second one, which would describe the product twice. The fix above applies in both cases.
  4. Search for priceValidUntil in the same page source. If your theme writes “tomorrow” or a nearby date on every page view, consider whether it is worth keeping: it matches no offer end.
  5. If your sales have a real end date, declare it in the Merchant Center feed with sale_price_effective_date, through your feed tool.

Limits

Three product pages per store give a picture of each store, not of its whole catalog. Clients of partner agencies are probably better maintained than average. A theme may have been modified by an agency: the tables describe the stores observed, not the theme as shipped. 155 pages shown in a currency other than the euro were excluded from the price check. We cannot see the stores’ Merchant Center feeds: layer 2 relies on Google’s documentation. The WooCommerce sample (10 sites with pages on sale) shows what the native code produces, not the state of the WooCommerce market.

Robustness check: the same day, the same measurement on the barometer’s June 2026 list (445 stores, 1,325 pages) leads to the same conclusions, with 0 strikethrough prices in the markup of 147 products on sale.

Kamil Kaderbay, founder of Verity Score. Study of October 13, 2026.

To go further: the Shopify product data that matters for AI Max and the Shopify GEO barometer.


To go further than the manual check: the Verity Score audit reads your product pages’ markup and compares the marked-up price with your Shopify data. It is available in the public Shopify app: install Verity Score →