Aizii Learn · AI Search
← Back to Learn

How to remove outdated product info from AI answers

You cannot edit an AI answer directly. There is no correction form that rewrites what an assistant says about your products for everyone who asks. The outdated fact lives somewhere the assistant reads, usually your own site, a retailer listing, a directory, or a search index that has not caught up with a change you already made. The fix runs in that order: find the exact wrong fact, find the page it came from, correct or remove that page, make sure the crawlers can read the change, ask the search engines to refresh the stale result, and retest the same prompt until the answer changes. An answer built from live search can change after the next crawl. An answer built from training data waits for a newer model, and no source fix moves it sooner.

Start by deciding which kind of outdated info you are dealing with. A price that changed is different from a product that no longer exists, and both are different from a sale that ended. A changed price needs the new price on the page and in the markup. A discontinued product needs the page removed or redirected, plus a clear statement of what replaced it. An ended sale needs the sale language taken off the page, not just the banner hidden.

Where it goes wrong

Outdated product info survives because the same fact lives in more than one place, and the assistant reads whichever copy it finds first.

  • The live page still shows the old fact. The price, the size range, or the availability on the product page was never updated, so every crawl brings back the same wrong answer. This is the most common cause, and the fix is on your own site.
  • The page was fixed, but the markup was not. The visible text says the new price and the structured data still carries the old one. Google says structured data must match the visible text, so a mismatch keeps the wrong number in circulation.
  • The product is gone, but the page is still live. A discontinued product keeps its page, its price, and its buy button, and assistants keep recommending it.
  • The page is gone, but the search result is not. You deleted the page, yet Google or Bing still shows the old snippet, because neither engine has recrawled the address since the deletion. The fix needs a refresh request, not another edit.
  • A retailer or marketplace still lists the old fact. Your site is correct, but a retailer page, a marketplace listing, or a review site still shows the old price or the discontinued model, and the assistant cites that page instead of yours.
  • The crawler cannot read the correction. The updated page sits behind a block in robots.txt, a firewall rule, or a login wall, so the assistant keeps reading the old copy it already has. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, and Perplexity says the same in effect for PerplexityBot, so a block on the search crawler is a block on the correction too.

The steps

Step 1. Capture the outdated answer exactly. Save the prompt word for word, the full answer, every citation or link shown, and which assistant gave it. Circle the exact wrong sentence: the old price, the discontinued model name, the ended sale. Ask the same prompt in a second assistant and save that answer too. You need the exact wording to trace the source, and you need the same wording later to test whether the fix worked.

Step 2. Trace the fact to its source page. Open every citation in the answer. If there are no citations, search the wrong fact as a phrase and see which pages come up. Note whether the source is your own site, a retailer, a marketplace, a directory, or a review site. Note the address of each page that carries the old fact. One wrong answer often has two or three source pages, and fixing only one of them leaves the answer unchanged.

Step 3. Fix every page you control. Correct the fact on the product page, the category page, the homepage if it names the product, and the FAQ if it repeats the price or the availability. Put the corrected fact in plain text a reader can see, not only in an image or a script. If the product is discontinued, say so in plain words on the page that replaces it, and name the product that replaces it. Update the structured data on the same page in the same edit, so the markup and the visible text agree.

Step 4. Remove or redirect a dead product page. If the product will not come back, take the page down or redirect it to the closest live page, such as the category page or the replacement product. Do not leave the old page live with an add to cart button that goes nowhere. Do not redirect every dead product to the homepage, since that tells neither the shopper nor the crawler what replaced it. Update the sitemap in the same change, so the dead address leaves the file the day it leaves the catalog.

Step 5. Fix the pages you do not control. Contact each retailer, marketplace, or directory that still shows the old fact, and send them the link to your corrected page plus the exact corrected sentence. Ask for the change in writing and keep a note of who you asked and when. Where a listing belongs to a marketplace and cannot be edited by you, use the marketplace support route for a catalog correction.

Step 6. Make sure the search crawlers can read the correction. Check your robots.txt file for the search crawlers: OAI-SearchBot for ChatGPT search, PerplexityBot for Perplexity, and Googlebot for Google. Confirm the corrected pages return a normal success response when those crawlers visit, and that no firewall or bot rule blocks them. OpenAI says it can take about 24 hours from a robots.txt update for its search systems to adjust, and Perplexity says the same for its systems, so test the robots file first and wait before you judge the result. For Google, use the URL Inspection tool in Search Console to inspect the corrected address and request indexing.

Step 7. Ask the search engines to refresh the stale result. For Google, use the Refresh Outdated Content tool. Google says to use it when content was deleted from a site but still comes up in search results, and that if you already updated a live page you do not have to fill out the form, since a recrawl will pick up the change. If the page no longer exists, submit its address and Google checks that it is gone. If the page still exists but changed, Google may ask for one or two words that appeared in the old snippet and no longer appear on the live page. For Bing, use the Bing Content Removal tool, which offers a page removal option for dead pages and an outdated cache option for pages that are still live. A refresh request only updates the search result. It does not remove a page from the web, and it does not change what an assistant built from training data says.

Step 8. Retest the same prompts and keep a log. Ask the same questions again with the same wording, in each assistant, and record the answer and the citations each time. Search based answers can move after a recrawl. Training based answers will not move until a newer model arrives, so keep the log and stop expecting a change between model releases. If the same wrong fact keeps appearing with the same citation, the source page is still wrong, or the crawler still cannot read the correction. Go back to step 2 and check the citation again instead of repeating the refresh request.

The machine-readable piece

This robots.txt snippet allows the search crawlers to read your corrected pages while keeping the training crawler out. The two settings are independent, as OpenAI documents.

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: GPTBot
Disallow: /

This Product markup shows a discontinued product as out of stock, with the visible page saying the same thing. The availability value comes from the schema.org vocabulary. No price is shown here, since the page no longer sells the product.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Example stand mixer, retired model",
  "url": "https://[yourdomain.com]/products/example-stand-mixer",
  "offers": {
    "@type": "Offer",
    "availability": "https://schema.org/OutOfStock"
  }
}
</script>

For a retired page you want out of Google search results entirely, add a noindex directive to the page itself and let the crawler fetch the page so it can read the directive. Google lists noindex among the controls for limiting what shows from a page in search. A noindex page that is also blocked in robots.txt cannot be read, so do not block the page and noindex it at the same time.

<meta name="robots" content="noindex">

What it looks like when it works

On the page, the change is plain to a human reader. The product page shows the current price in normal text, or it says the product is discontinued and names the replacement. The category page no longer lists the retired model. The sitemap no longer contains the dead address. The structured data on each page says the same thing as the visible text. An assistant asked about the product gives the current price, or says the model is discontinued and names the replacement, and the citation points at the corrected page.

Say you run a kitchenware store and you retired a stand mixer model after replacing it with a newer one. A shopper asks an assistant for your best stand mixer, and it names the retired model at the old price, citing your old product page and a retailer listing. You take the old product page down and redirect it to the new model page, which says in plain text that the old model is discontinued and names the new one. You update the sitemap the same day, ask the retailer to update its listing, and check that OAI-SearchBot and PerplexityBot can fetch the new page. You submit the old address to the Refresh Outdated Content tool, and you request indexing of the new page in Search Console. You retest the same prompt and log the answer. The search based answer changes after the next crawl and names the new 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 removing outdated product info from AI answers by fixing the source and asking search engines to refresh. It does not cover correcting a wrong fact that was never true, such as a made up specification, which is a different job with its own guide. It does not cover legal removal of defamatory or personal information, which uses separate request routes. It does not cover the monitoring work of catching the next outdated answer early, or the full technical audit that checks every crawler and every page.

Check your own store now

Run the free scan. One click, no signup, and it shows whether AI crawlers can read your store 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.