Aizii Learn · AI Search
← Back to Learn

Why does ChatGPT show outdated prices for my store

ChatGPT quotes the price from the copy it last read, and that copy is often not your current product page. It may be a cached search result, an old sale page that is still live, a reseller listing, or the structured data on your own page that still holds the old figure after the visible price changed. A price change on the product page alone does not reach ChatGPT. Every copy of that price has to change, and the search crawler has to come back and read the corrected page, before the answer moves.

OpenAI's own help pages warn that search results and citations can be incomplete, outdated, or incorrect. Treat a quoted price as a lead to the source it came from, not as proof that your site is wrong.

Why the price in the answer is stale

Outdated prices come from a short list of causes, and each one points to a different fix:

  1. A stale indexed copy. OAI-SearchBot last read the page before the price changed, and it has not returned since. Crawlers revisit pages on their own schedule, and a robots.txt change takes about 24 hours to reach OpenAI's search systems, so a fix does not show up in answers the same hour it goes live.
  2. Markup and page disagree. The visible price changed, but the JSON-LD on the same page still carries the old price, or the other way around. Google's guidance for AI features says structured data should match the visible text on the page. When the two disagree, an assistant can quote either one, and the wrong one is usually the one you forgot to update.
  3. An expired sale still looks live. A promotion ended, but the sale page is still published, or the markup still carries a sale price with an end date that has passed. Google's product documentation says a listing may not display when the price end date in the markup indicates a past date, which is a sign the markup is stale, not a live offer.
  4. The answer cites the wrong source. A reseller, a marketplace, a coupon site, or an old blog post still shows the old price, and ChatGPT cites that page because it is the copy the search index holds. Fixing your own page does nothing until that cited source changes or your page becomes the stronger source.
  5. The answer comes from training data. A price remembered from training changes only when a newer model arrives. No source fix moves it right away, which is why the test in step 4 below separates search answers from remembered ones.

Step 1: Find the source ChatGPT actually quoted

Ask ChatGPT the same question a customer would ask, with web search on, and open every citation it shows. Write down the exact price, the page it came from, and whether the page is yours, a reseller, or a third-party site. If the answer shows no citations and reads like remembered knowledge, treat it as training-based and skip to step 5 for what that means. Do not start editing until you know which copy is being quoted, because the wrong fix leaves the quoted page untouched.

Step 2: Change every copy of the price at once

Make one change that covers all three places a price lives:

  • The visible price on the product page, in plain text a shopper can read.
  • The structured data on the same page, with the same price and the same currency code, and the same availability value as the page shows.
  • The sale end date in the markup. If the price is a sale price, set its end date to the real end of the sale, and update it when the sale is extended. If the sale is over, remove the sale price from the markup instead of leaving it with a past end date.

Then check the places outside the product page: the category page, the homepage feature, the FAQ, and any banner that still shows the old figure. A price that is correct on the product page and wrong on the category page is still a conflicting source.

Step 3: Clear away the old pages

Remove or redirect any page that exists only to show the old price: ended sale pages, old promotion landing pages, and duplicate product pages from a previous catalog. If a page must stay live for another reason, update the price on it or remove the price from it. A live page with an old price is a live source for the next answer, and ChatGPT has no way to know you meant to retire it.

Check the resellers and marketplaces that carry your products next. Where a listing belongs to someone else, ask the site owner to update it and send them the link to your corrected product page. Where the listing is yours, update it the same day you update your site, so the two do not drift apart again.

Step 4: Get the corrected page recrawled

Confirm OAI-SearchBot is allowed in robots.txt and not blocked at the edge, because a corrected page the search crawler cannot fetch will never replace the stale copy. Check your logs for recent OAI-SearchBot visits to the product page. Then test with a direct fetch: ask ChatGPT to read the product page URL itself. That uses ChatGPT-User, the per-request fetcher, which is not the search index, but it tells you whether the live page now shows the right price to a reader. Retest the original shopping question after the next crawl and log the price and the citations each time. Search-based answers can move after a recrawl. Training-based answers will not move until a model update, so keep the log and stop expecting a change between model releases.

Step 5: Know which answers you cannot move yet

If the same wrong price appears with no citations in a fresh chat, it is likely coming from training data. Fix the sources anyway, because the next model trains on the corrected web, but do not keep re-editing the page in the hope that the remembered answer will change this week. It will not. The work that pays off now is the search-based answer, which is also the answer a customer sees when ChatGPT searches the web for current prices.

A hypothetical, to make it concrete

Say you run an outdoor gear store and raise the price of a tent. You update the product page, but the JSON-LD still shows the old price, the category page still shows the old price, and the ended sale page for the tent is still live. A customer asks ChatGPT for the price, and it quotes the old figure, citing the sale page. The fix runs in this order: update the visible price, the markup, and the sale end date together, remove the ended sale page, and correct the category page. Then confirm OAI-SearchBot can fetch the product page, and retest the same question after the next crawl. The search answer changes. Any training-based version of the price waits for a newer model.

The fundamentals

These mechanics are shared across every engine and covered in full in their own guides:

What this guide does not cover

This guide covers a price that is wrong in ChatGPT's answers. It does not cover prices that are wrong in Google Shopping, in a marketplace feed, or in your own checkout, and it does not cover the monitoring work of catching the next price drift early. If a wrong price has already cost a customer money, fix the listing and handle the customer directly rather than treating a feedback button as the remedy.

Check your own brand now

Run the free scan. It takes seconds, no signup, and shows whether AI crawlers can read your corrected pages today: aizii.ai/aeo-scan

Run the free scan

Sources

Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.