How Do AI Assistants Handle a Product You Discontinued?

Why do AI assistants keep recommending a product you discontinued?

They keep recommending it because the evidence that it exists outweighs the evidence that it ended. Old reviews, comparison articles, and your own archived pages still describe the product as available. Unless you publish a clear, findable retirement notice and update the pages that mention it, assistants have little reason to change their answer.

This catches a lot of SaaS teams off guard. They retire a plan, a product line, or an add-on, update the pricing page, and move on. Months later a prospect arrives asking about the old product, because ChatGPT or Perplexity described it as a current option.

The fix is not a single page. It is a small, deliberate cleanup across the places where the old product still lives, starting with the ones you control.

Where do AI assistants get their picture of your product line?

They draw on two sources: what a model absorbed during training, and what a search-enabled assistant retrieves at the moment of the question. Training data reflects the web as it was at some past cutoff. Retrieval reflects what ranks and gets fetched today. A discontinued product can survive in both.

The training side is the slower one. If your product was widely written about before a model's cutoff, the model may remember it as a current offering. You cannot edit that memory directly. What you can do is make sure that when an assistant searches, the freshest and clearest pages tell the true story.

The retrieval side is where you have leverage. Assistants such as ChatGPT with search, Perplexity, Gemini, and Google AI Overviews pull pages that seem relevant and authoritative for the question. If the most relevant page about your retired product is an old feature page that still reads as current, that page becomes the answer.

Why does a retirement announcement alone not fix it?

A single announcement is one page arguing against many. If your blog post says the product ended, but your old feature page, three review sites, and a dozen comparison articles still describe it as available, the weight of evidence still says it exists. Assistants tend to reflect the consensus of what they find, not the single newest post.

Announcements also tend to be written for existing customers. They say things like "this plan is being sunset, here is how to migrate." That is helpful for a customer, but a buyer's question to an assistant is usually different. They ask what the product does or how much it costs. Your announcement may not match that question at all.

So the announcement is necessary but not enough. Treat it as the anchor, then make every other page you control agree with it.

What should the retired product's own page say?

Keep the page live and rewrite it as a clear retirement notice. Say in the first sentence that the product is discontinued, when it ended, and what replaced it. Then link to the replacement and the migration guide. Deleting the page leaves old links and old citations pointing nowhere, which can make the stale story stronger.

The first sentence matters most because answer engines tend to quote short, direct passages. A sentence like "Product X was discontinued in March 2026 and replaced by Product Y" is easy to extract and hard to misread. Bury that fact in the fourth paragraph and an assistant may quote the old feature description above it instead.

Update the page title and meta description too. A title that still says "Product X: features and pricing" signals a live product. Change it to something like "Product X has been discontinued: what replaced it." That one edit changes what search results and assistants see before they even open the page.

Should you redirect or keep the old URL?

Keep the URL and change the content, at least for the first several months. A redirect sends people to the replacement but erases the explicit statement that the old product ended. An assistant that fetches the old URL then sees only the new product, and may blend the two into a confusing answer.

There is a case for a redirect later. Once the old product stops coming up in sales calls and search queries for it fade, redirecting the retirement page to the replacement is a reasonable cleanup. By then the retirement notice has done its job.

If your team already redirected the page, you can still recover. Publish a short, dedicated page that states the retirement plainly, link to it from the replacement page, and include it in your sitemap. You need one page that says the product ended in words an assistant can quote.

Which other pages on your site still describe the old product?

Search your own site for the product name. You will usually find it on the pricing page footnotes, comparison pages, integration pages, older blog posts, help docs, and customer stories. Each mention that reads as current is a vote for the old answer. Update them, or add a one-line note that the product has been retired.

Help docs are the most overlooked. They often rank well and get retrieved because they are specific and detailed. An old setup guide for a retired feature can easily become the page an assistant cites. Add a notice at the top of each one, or move them to an archive section with a clear label.

Customer stories are the trickiest. You do not want to delete proof. Add a short line noting the customer used a product that has since been replaced, and if possible, a note on what they use now. That keeps the story honest without pretending the old product still exists.

What about review sites and articles you do not control?

Ask for updates where you have a relationship, and leave the rest alone. If you hold a vendor profile on review platforms like G2 or Capterra, check their own help docs for how to update or retire a listing. For articles by others, a polite note to the author with your retirement page sometimes works. Skip the rest.

Focus on the third-party pages that actually rank for questions about the old product. Search the product name in a normal search engine and in an assistant, and note which outside pages appear. Those are the ones worth a request. A page that never surfaces is not shaping anyone's answer.

If you have structured data on your own pages, keep it accurate too. A Product schema block still describing the retired item as available sends a conflicting machine-readable signal. Update or remove it so the markup agrees with the visible text.

How do you check whether assistants have caught up?

Ask them directly, on a schedule. Each month, ask a few assistants the questions a buyer would ask, such as what the old product does, what it costs, and whether it is still available. Record the answers and which sources they cite. When the cited sources switch to your retirement page or replacement, the cleanup is working.

Keep the questions consistent so you can compare month to month. I use a short sheet with the question, the assistant, the date, the answer in one line, and the cited URLs. It takes twenty minutes and turns a vague worry into a trend you can see.

Expect the retrieval-based answers to change first and the purely model-based answers to lag. That is normal. Your job is to make the live evidence clean. The rest follows as models are refreshed.

How is this different from a rebrand or a name change?

A rebrand keeps the product and changes its name. A discontinuation ends the product. With a rebrand, you want assistants to connect old and new names as the same thing. With a discontinuation, you want them to stop recommending the old thing entirely and point to its replacement instead. The tactics overlap, but the goal is opposite.

I covered the rebrand case in how AI search handles a company rebrand, and a lot of the same mechanics apply here. Both depend on clear first sentences, consistent pages, and patience.

If you are retiring a feature rather than a whole product, the approach is lighter but similar. My notes on announcing a retired feature to SaaS customers cover the customer side, which you should handle before the public cleanup.

What should you do next?

Search your site for the retired product's name and list every page that mentions it. Rewrite the product's own page as a retirement notice with a direct first sentence, update help docs and comparison pages, and fix any structured data. Then start a monthly check of what assistants say, and watch the citations move.

The underlying lesson is that AI assistants reflect the web you leave behind. If you want them to tell the current story, the current story has to be the clearest one on the page.

If you are retiring a product and want help planning the cleanup across your Webflow site, reach out. Let's chat about which pages carry the most weight and where to start.