AI Agents Can Now See Your Sale Price

On October 9, 2026, Shopify started sending one more field from your product admin to AI shopping agents: the compare-at price. If a shopper asks ChatGPT, Copilot, or an agentic storefront whether your hoodie is on sale, the agent no longer sees only the current price. It can see the number you would strike through on your own product page — and it may repeat the discount claim that pair of numbers makes.

That is the whole story, and it is a listing story. The compare-at price is a field you edit in the Shopify admin, per product, per variant. It is not schema markup, not Search Console, not a developer integration. If you sell on Shopify, the decision — what agents see when they read your pricing — now sits inside your product data, and this article walks through exactly what changed, what it means for discount claims you may have set years ago, and what to inspect this week.

What Shopify shipped on October 9

Three facts from Shopify’s own changelog and help documentation define the change.

First, the syndication. “Shopify shares your compare-at price with AI shopping agents through Shopify Catalog, so AI agents can see when a product is marked down,” the changelog entry reads. “Previously, AI agents reading Catalog only saw the current price and markdown prices displayed in storefront were not shown.” In other words: for the entire life of agentic storefronts until last week, your storefront showed a strike-through price while agents reading the structured feed saw nothing but the sale number. That asymmetry is gone.

Second, the toggle. Merchants control the behavior in the Shopify admin: Sales channels > Agentic > Sources > Shopify Catalog > Review product data mapping > the “Compare-at” section. Shopify’s help documentation describes the semantics precisely — when the setting is on, Catalog sends a compare-at price only if it is higher than the product’s current price, and “AI agents might use it to advertise strikethrough pricing.” When it is off, Catalog sends only the current price. Shopify says each store “starts with a default selection” but does not publish which one, so treat your own admin as the source of truth rather than any summary you read elsewhere — including this one.

Third, the plumbing. The same help doc notes that Shopify Catalog sends an eligible compare-at price through the list_price field in the Universal Commerce Protocol (UCP) Variant specification. The UCP schema, published in the protocol’s open repository, defines list_price as “List price before discounts (for strikethrough display)” — a first-class field sitting beside price, the current selling price, in the same variant object every participating agent reads. The preview tab in Catalog Mapping even shows the rendered result: the compare-at price struck through next to the current price, with a badge showing the percentage discount.

A field that used to be page decoration is now machine-readable pricing context, shipped through the same structured channel — Shopify Catalog — that we have covered as the backbone of GEO for marketplace sellers.

Why two numbers are better than one — when both are true

Why would Shopify bother? Because a single price answers “what does it cost” and nothing else. Two prices answer “is this a deal,” and deal-seeking is a large share of what shoppers ask AI shopping systems to do.

There is also a hard observation about price quality on AI surfaces. Over 23 days (August 9–31, 2026), ProductRise ran the same shopping queries through Google AI Mode and traditional Google search at the same moment — more than 100,000 result pages covering more than 2 million product listings. Among products that appeared on both surfaces, prices disagreed 38.1% of the time; when they disagreed, AI Mode showed the higher price 68.4% of the time, an average premium of 21.6% on matched products. Most of those disagreements are stale or mismatched feed data, not fraud — but they show what happens when a surface reads a cached number instead of a live one.

Shopify Catalog’s continuous sync is designed to prevent exactly that: the agentic storefronts documentation says Catalog “continuously updates your product data, ensuring that there’s accurate inventory and pricing across every AI channel.” The compare-at field extends the same guarantee to your markdown signal. If your sale is real and your data is current, a syndicated compare-at price is a recommendation asset: the agent can present the shopper with a genuine discount rather than a bare number it cannot contextualize.

This is our read of the incentive, clearly labeled as inference. Neither Shopify nor any agent platform has published conversion data on strikethrough display in AI answers, and no one can promise that sending a compare-at price produces a sale, a citation, or a placement. What the documentation does promise is narrower and more useful: the signal now travels.

The honesty gate your compare-at column just inherited

Here is the part most coverage of this change skips: whatever is sitting in your compare-at column now has an audience, and markdown claims have a legal shape.

The FTC’s Guides Against Deceptive Pricing, at 16 CFR 233.1, draw the line on former-price comparisons. A former price is a legitimate basis for a bargain claim if it is “the actual, bona fide price at which the article was offered to the public on a regular basis for a reasonably substantial period of time.” A price that was “artificial, inflated” just “for the purpose of enabling the subsequent offer of a large reduction” is fictitious, and the guides’ own example is a retailer who briefly marks a $7.50 pen up to $10 so he can advertise “Were $10, Now Only $7.50.” The guides also note that no sales at the former price are required — but the price must have been openly and actively offered, in good faith, for a reasonably substantial period.

Now map that onto a typical Shopify catalog. Compare-at values usually come from one of three places: a price you actually charged for a real season, a supplier MSRP, or a number someone typed to make the discount look bigger. The first two are defensible. The third was always a liability on your product page — and as of October 9 it is a liability an AI agent can quote to a shopper who never visits your site. An inflated markdown claim, repeated by a trusted assistant with your product attached, is a chargeback and a trust problem you manufactured yourself. This is our recommendation, not a legal conclusion; if your catalog is complex, ask a lawyer.

One more caveat worth stating plainly: if you sell under supplier minimum-advertised-price agreements, whether an agent quoting your compare-at price counts as “advertising” under a given MAP policy is a question Shopify’s documentation does not answer. Get the answer from each supplier in writing before BFCM season, not after a violation notice.

What to inspect this week

The audit is an afternoon, and it happens in your product admin and Catalog Mapping — not in Search Console. This is the same field-level operating method we teach in Shopify Product GEO, applied to pricing.

1. Check your toggle state. Open Sales channels > Agentic > Sources > Shopify Catalog > Review product data mapping and look at the “Compare-at” section. Shopify does not publish the default, so record what your store actually shows before you assume anything.
2. Export and sort your compare-at column. In Products, export a CSV and filter for rows where compare-at is greater than price. Sort by the size of the gap. Note that Catalog only sends the field when compare-at is higher than price — a compare-at that is lower or equal simply does not travel.
3. Tie each flagged value to a document. For every large markdown: a dated price you charged (order history or price-change log), or a supplier MSRP sheet. If you can tie it to neither, clear the field. Under the FTC guides, “nobody remembers why it’s $195” is not a basis for a bargain claim.
4. Check your feed tools’ field mapping. If supplier inventory or sync apps write MSRP into compare-at automatically, your store has been claiming markdowns nobody decided on. Fix the mapping at the source, then re-check five products before re-running a full import.
5. Decide the toggle deliberately, per the three buckets. Every value documented: leave it on — a real markdown is a genuine signal. Mostly documented: finish the cleanup this week, or switch it off until you do. Mostly undocumented: switch it off, audit, then re-enable. Turning the toggle off never removes your products from AI channels; Catalog keeps sending the current price, title, media, and availability exactly as before.
6. Re-check before scheduled sales. If you use Rollouts or scheduled discounts for BFCM, set compare-at prices on the products before the sale activates. Agents read Catalog continuously, so the discount signal needs to exist when the traffic arrives, not the morning after.

For apparel sellers this lands hardest around seasonal transitions — the winter coat that was $220 in February and is $138 in October is the textbook honest markdown, provided February’s price was real. That is the entire game: the strikethrough must be a fact, because a fact is now what gets syndicated.

How FirstShelf can help

FirstShelf audits the exact fields AI shopping systems read when they price, filter, and recommend products — including the compare-at column, its consistency against your current prices, and the toggle state that decides what travels. We score your listings the way agents parse them and return a prioritized fix list: which markdown claims are documented, which fields are feeding agents stale or inflated numbers, and what to correct before your biggest traffic weeks. That is listing work, not schema cleanup, and it is the difference between an agent that can vouch for your discount and one that silently quotes a number you cannot defend.

Questions sellers are asking

Does the compare-at toggle control whether my products appear in AI answers?
No. The toggle governs one field. Your products stay in Shopify Catalog and continue to be discoverable by AI channels with their current price, title, media, and availability whether the toggle is on or off.

What happens if my compare-at price is lower than or equal to my price?
Nothing is sent. Shopify’s help documentation states Catalog sends a compare-at price only if it is higher than the current price; otherwise the agent sees the current price alone.

I never set compare-at prices. Do I need to do anything?
Only decide whether you want the signal. A product with no compare-at value sends no markdown claim. If you run sales by lowering price alone, you can set an honest compare-at — the price you actually charged before — to make the discount legible to agents.

Is showing a compare-at price to an AI agent a MAP violation?
It depends on the supplier agreement. Shopify’s documentation does not address how agent displays interact with minimum-advertised-price terms, so ask each supplier in writing before your next promotion cycle.