# The Price Has an Expiry Time: GEO for Volatile Commerce Facts

> A product price is not a timeless fact. It is a value attached to a currency, variant, seller, quantity, audience, market, and period of validity.

- Canonical: https://www.aixindar.com/news/the-price-has-an-expiry-time-geo-for-volatile-commerce-facts
- Markdown: https://www.aixindar.com/news/the-price-has-an-expiry-time-geo-for-volatile-commerce-facts.md
- Author: Daoyu Guan — https://www.aixindar.com/experts/daoyu-guan
- Published: 2026-09-28T12:03:47.504Z
- Last updated: 2026-09-28T12:03:47.619Z
- Evidence checked: Not separately recorded in CMS
- Editorial status: Published
- Corrections: No correction record supplied by CMS.

A product price is not a timeless fact. It is a value attached to a currency, variant, seller, quantity, audience, market, and period of validity. Google's September 23, 2026 documentation update added sale-pricing support through the `priceType` property in merchant-listing structured data, while its holiday commerce update urged retailers to keep product feeds accurate for AI discovery. The GEO lesson is larger than one schema field: prices, stock, promotions, delivery dates, and loyalty benefits need explicit scope and expiry. When pages, markup, feeds, and checkout disagree, an AI answer may repeat a value that was technically published but commercially false for the user's situation.

## The difference between a number and an offer

"$79" looks like a fact. In commerce systems, it is incomplete. A usable offer might be represented as:

`Offer = product variant + seller + amount + currency + eligibility + geography + quantity + start time + end time + availability`

Remove any of those dimensions and the same number can mislead. The $79 price may apply only to a refurbished unit, a loyalty tier, a two-pack, a U.S. store, or a weekend promotion. It may exclude tax, shipping, installation, or a required subscription.

Conversational search increases the risk because the answer often compresses a product page into one sentence. "The camera costs $79" is much easier to remember than "the 64 GB model was advertised at $79 to Gold members in the U.S. until Sunday." GEO for commerce must preserve the conditions that make the claim true.

## What Google's sale-pricing update actually defines

Google's merchant-listing documentation distinguishes three price classes:

- the **active price**, which is the amount currently offered;
- the **strikethrough price**, which is the higher regular price shown during a sale;
- the **member price**, which is available to a defined loyalty tier.

For sale pricing, the current amount remains the active price. The original amount is represented in a `UnitPriceSpecification` with `priceType` set to `https://schema.org/StrikethroughPrice`. Google's documentation warns against marking the active sale price with `priceType`. It also says `priceType` and `validForMemberTier` must not be used together on the same price specification.

The September 23 change improves the machine-readable distinction between "what the customer pays now" and "the reference price that is crossed out." It does not validate whether the reference price is lawful, genuine, or commonly charged. That remains the merchant's responsibility under applicable consumer law and platform policy.

## Time belongs inside the claim

Google recommends using `validFrom` and `validThrough` or `priceValidUntil` to describe a sale period, with time and time-zone information when possible. This matters at boundaries. A promotion ending at midnight in London ends eight hours later than midnight in Shanghai and three hours earlier than midnight in New York on many dates.

For an international retailer, "Ends Friday" is not machine-safe. A more precise record is:

- start: `2026-11-20T08:00:00-05:00`
- end: `2026-11-30T23:59:59-05:00`
- market: United States
- currency: USD
- eligible variant: SKU-13-BLK-512
- audience: all customers

The visible page can translate that record into natural language. The source system should retain the exact interval.

Do not keep expired sale markup on the page because a campaign banner was removed. A stale `priceValidUntil` can make a listing ineligible or expose a price that checkout will reject.

## Four surfaces must agree

Commerce facts commonly travel through four systems:


| Surface          | Purpose                        | Failure example                                         |
| ---------------- | ------------------------------ | ------------------------------------------------------- |
| Product page     | Explains the offer to a buyer  | Banner shows $79 but selected variant is $99            |
| Structured data  | Describes the page to machines | Markup retains last week's sale                         |
| Merchant feed    | Supplies catalog data at scale | Feed updates once daily during an hourly promotion      |
| Cart or checkout | Executes the transaction       | Coupon or membership condition appears only at checkout |


An AI answer may draw from the page, feed, an indexed copy, an affiliate, or a shopping integration. Updating one surface is not enough. The most reliable design has one offer service or catalog record that publishes to all four, with validation at each destination.

If the systems must remain separate, create a reconciliation job that compares the decisive fields: SKU, seller, currency, active price, reference price, membership condition, availability, and expiry. Treat a critical mismatch as an incident, not a cosmetic warning.

## The source of truth needs event history

A current-value database answers "what is the price now?" It cannot always answer "what price did the AI quote yesterday, and was it valid then?" Debugging requires history.

Keep an append-only event record for offer changes:


| Event field    | Example                       |
| -------------- | ----------------------------- |
| Offer ID       | `offer-us-q13-512-black`      |
| Event time     | `2026-09-27T14:05:22Z`        |
| Effective time | `2026-09-28T04:00:00Z`        |
| Previous value | USD 109.00                    |
| New value      | USD 89.00                     |
| Reason         | Holiday campaign H26-04       |
| Actor          | Pricing service account       |
| End time       | `2026-10-02T03:59:59Z`        |
| Destinations   | Page, JSON-LD, feed, checkout |


Event time and effective time are different. A price can be entered today but begin next week. Without both, teams cannot determine whether a crawler saw an error or a future offer published too early.

## Member prices need explicit eligibility

A loyalty price is not the universal product price. Google's documentation represents it with `validForMemberTier`, pointing to a defined program tier. The visible page should also state the condition before the customer commits to a purchase path.

For AI answers, eligibility should travel with the amount:

> Gold members can buy the 512 GB variant for $89 through October 1; the public price is $109.

That sentence is longer than "$89 sale price," but it prevents three errors: applying the member price to everyone, attaching it to the wrong variant, and presenting an expired amount.

If a membership can be joined immediately for free, say so. If it requires a paid subscription, prior purchase, geographic residency, or approval, include that condition. A hidden eligibility rule turns an attractive citation into a trust failure at checkout.

## Reference prices are evidence claims

A strikethrough amount implies that the active price is lower than a meaningful regular price. That implication should be supportable. Merely loading a high price for a short period before a sale can create legal and reputational risk. Requirements vary by jurisdiction, so commerce teams should use local legal guidance rather than a global marketing convention.

From an E-E-A-T perspective, keep records that answer:

- Where and for how long was the reference price offered?
- Which seller and market used it?
- Does it apply to the same SKU and condition?
- Was the product actually available at that amount?
- Which promotion authorized the lower price?

Do not ask content writers to invent urgency. The pricing system should provide the claim and its evidence; the page should explain it accurately.

## AI discovery raises the cost of feed drift

Google's September 16 holiday-commerce announcement said AI performance insights were becoming available in several markets and emphasized product-feed accuracy. It also described product feeds as inputs to conversational shopping and business-agent experiences. OpenAI's March 2026 shopping announcement similarly described merchant feeds as part of product discovery in ChatGPT through the Agentic Commerce Protocol.

These are separate ecosystems with different controls. The common operational lesson is that product facts can enter AI answers through structured catalog channels rather than a fresh read of the public page. A merchant should map every outbound feed and integration, then decide which system owns corrections and takedowns.

An obsolete affiliate feed can outlive the campaign that created it. A marketplace can cache the old amount. A translated landing page can show a conversion based on yesterday's exchange rate. The retailer's own page may be correct while the wider evidence graph remains wrong.

## A promotion launch checklist

1. **Freeze the offer definition.** Confirm SKU, market, seller, currency, audience, quantity, start, end, stock rules, and exclusions.
2. **Validate the reference price.** Retain the evidence required by the relevant jurisdiction and policy.
3. **Publish from one approved record.** Update page, structured data, feed, ad, agent knowledge, and checkout from the same fields where possible.
4. **Test boundary times.** Check before launch, at launch, one hour before expiry, at expiry, and after rollback in every supported time zone.
5. **Test variants.** Select different sizes, colors, conditions, and sellers. Confirm that the visible and marked-up price follows the selected object.
6. **Test anonymous and member states.** Verify what a non-member sees and what a qualifying member pays.
7. **Capture evidence.** Save page HTML, structured data, feed record, cart total, timestamp, and market.
8. **Remove or expire every surface.** Treat campaign closure as a deployment with its own acceptance test.

## Diagnose a wrong AI price

When an AI answer reports the wrong amount, do not immediately rewrite the product description. Trace the claim.

1. Record the exact prompt, answer, source links, market, time, account state, and selected variant.
2. Determine whether the amount was ever valid for any seller, tier, or region.
3. Inspect the canonical product page and its raw structured data.
4. Inspect current and historical feed records.
5. Check marketplaces, affiliates, PDFs, emails, and translated pages that may expose the amount.
6. Correct the authoritative source first, then downstream channels.
7. Request recrawl or refresh only through documented mechanisms.
8. Retest after the relevant discovery or cache interval.

The result may be "correct price, wrong membership scope" or "expired affiliate feed," not simply "AI hallucination." The distinction determines the fix.

## Measure factual commerce quality

Useful metrics include:

- **critical-field agreement:** percentage of offers whose page, markup, feed, and checkout match;
- **promotion boundary accuracy:** passes at start and expiry times;
- **stale-offer half-life:** time until an expired amount disappears from monitored surfaces;
- **eligibility retention:** percentage of AI answers that preserve membership or market conditions;
- **variant precision:** percentage of answers tied to the correct SKU;
- **checkout truth rate:** quoted offers that can actually be completed under stated conditions;
- **correction latency:** time from source fix to verified external change.

Share of voice is not enough. A prominent wrong price can be worse than no mention.

## Frequently asked questions

### Should the active sale price use `priceType`?

Google says no. The active price remains the current offer price; the higher reference price uses `StrikethroughPrice`.

### Can one price specification be both a member price and a strikethrough price?

Google's documentation says `priceType` and `validForMemberTier` must not be used together on the same price specification.

### Is structured data more authoritative than the visible page?

It should describe the visible offer, not contradict it. Google may also use feed data, and the transaction is ultimately tested at checkout.

### How often should prices be refreshed?

Match the refresh interval to the rate of change and commercial risk. Flash sales and availability require a faster path than a stable list price.

### Does accurate markup guarantee inclusion in AI shopping results?

1. It improves the clarity and consistency of product data but does not guarantee eligibility, selection, ranking, citation, or sale.

## Sources and evidence boundary

- [Google Search Central, "Merchant listing structured data," accessed September 28, 2026](https://developers.google.com/search/docs/appearance/structured-data/merchant-listing)
- [Google Search Central, documentation updates for September 2026](https://developers.google.com/search/updates)
- [Google, "Boost your holiday sales with these agentic commerce updates," September 16, 2026](https://blog.google/products-and-platforms/products/shopping/google-shopping-updates-holiday-shopping/)
- [Google Search Central, "Introduction to Product structured data," accessed September 28, 2026](https://developers.google.com/search/docs/appearance/structured-data/product)
- [OpenAI, "Powering Product Discovery in ChatGPT," March 24, 2026](https://openai.com/index/powering-product-discovery-in-chatgpt/)

Platform documentation defines supported data fields and product behavior, not legal compliance in every market. The governance model and metrics above are operational recommendations. Merchants should verify local pricing, promotion, tax, and consumer-protection obligations with qualified counsel.

## Editorial references

- [Editorial policy](https://www.aixindar.com/editorial-policy)
- [Research methodology](https://www.aixindar.com/research-methodology)
- [Corrections policy](https://www.aixindar.com/corrections)