What to Preserve During a Site Migration When Other Sites and AI Answers Cite You

During a website move, links from other sites and citations in AI-generated answers can keep pointing to old URLs. A URL is the specific web address of a page or file. If that address stops working, visitors who follow an old citation can lose the information they expected to find.

Before launch, identify the old URLs that people and AI answer products already cite. Keep their content available at those addresses where practical; otherwise, send each old URL to the closest equivalent new page with a permanent redirect. A redirect gives visitors a route to the replacement, but it does not guarantee that search engines or AI products will update their citations or preserve visibility.

Keep cited URLs and their content connected

A link from another site points to a particular URL. An AI answer’s citation can also link to a particular page. When you move that page, a permanent redirect, such as an HTTP 301 or 308 response, tells browsers and search engines that the address has changed for good. Google uses a permanent redirect as a signal that the new URL represents the content. Google also says that 301 redirects do not reduce PageRank. That guidance describes Google Search’s handling of link signals; it does not promise rankings or citation changes across every search or AI product. See Google’s redirect guidance.

Map every important old URL to the most relevant new destination. If an old article has a direct replacement, redirect to that article rather than the new homepage. If several pages are combined, make sure the destination still contains the information each old page was cited for. When no suitable replacement exists, return a genuine 404 or 410 response rather than sending visitors to unrelated content. Google warns that redirects to unrelated pages can be treated as soft 404s, meaning the destination does not appear to replace the missing page.

Keep the information from the old page available at its old address or on its replacement. Keep the answer, supporting details, and relevant images or downloads available. Include PDFs and other files in the URL inventory when outside sites link to them; their addresses may need their own redirects. Google’s site-move guidance specifically calls out images and downloads as content to account for during a move.

AI products use their own methods to find web pages and choose sources for answers. As of November 11, 2026, OpenAI documents OAI-SearchBot as a crawler that finds websites for ChatGPT search and GPTBot as a crawler that may collect data for training. Those crawlers have different purposes, and allowing search access makes a site eligible for inclusion without guaranteeing placement, according to OpenAI’s crawler documentation. A redirect can help a person or an automated program that visits web pages reach the replacement when it follows the old URL. It does not establish that an AI product will revisit the old citation, change its displayed link, or cite the new URL.

Find the URLs that already matter

Build your migration list from more than the current sitemap, which lists URLs on your site. Combine known URLs from the sitemap with landing-page data from analytics, Search Console data, server logs, and available backlink reports. Older articles, alternate URL formats, PDFs, and images can be absent from a current sitemap while still receiving visits or links.

Next, check which of your pages AI answer products cite. Search the AI answer products your audience uses for your brand. Also search for distinctive page titles and phrases from important pages. For each result, record the product, question, and date. Also note the cited URL and where the link leads. Repeat important checks after launch: answers and cited sources can vary, and a handful of observations cannot establish every citation in circulation.

Prioritize URLs with evidence of use, such as inbound visits, valuable links from other sites, or an AI answer citation you have observed. For each one, record the new destination and whether it will preserve the material the old page supported. A URL appearing in a backlink report or being reachable in a browser does not prove that an AI product cites or uses it.

Make the new site point consistently to the replacement

At launch, make the new URL the clear destination throughout the site. Update internal links and sitemap entries, and set the page’s canonical tag, which identifies the preferred URL when similar addresses exist, to the new address. Update language or regional page annotations if the site uses them. These signals help search engines understand the new URL structure, but a sitemap is a discovery hint, not a guarantee of crawling or indexing.

Remove temporary launch restrictions from the new pages. Check robots.txt and page-level noindex directives. Check whether pages require a login. Review firewalls and bot-protection rules. robots.txt tells crawlers which paths they may request; it does not reliably prevent a URL from appearing in Google’s search results. A crawler also needs to access a page to read a noindex instruction. Google explains this distinction in its robots.txt guidance.

After launch, test old URLs from outside a logged-in browser session. Confirm that each important old address reaches the intended page directly, that the destination returns successfully, and that it contains the expected content. Check server, CDN, or firewall logs for blocked requests and errors as well. A successful request from your browser or a command-line client shows only what that client received at that time. It does not prove that a particular crawler can access the page, that a search engine indexed it, or that an AI product cited it.

If you changed domains or subdomains, verify the old and new properties in Google Search Console, submit the new sitemap, and use Google’s Change of Address tool where it applies. Google says that tool is for domain or subdomain moves, not URL-path changes or HTTP-to-HTTPS changes. Its migration guidance recommends keeping redirects as long as possible, generally for at least a year. The recommendation supports Google’s recrawling and processing; it is not a guarantee that an AI citation will persist.

Watch for failures without overreading the results

The most damaging redirect mistake is sending every old URL to the homepage. That route may leave a visitor without the referenced answer, and Google warns that bulk redirects to unrelated destinations can be treated as soft 404s. Another common failure is removing the old domain or its redirects too soon. External pages and saved citations can continue to use old URLs after the new site launches.

Keep the redirects in place for as long as practical. Google recommends at least a year in its current migration guidance. Bing’s 2020 migration guidance recommends keeping old-domain redirects for at least one to two years, preferably longer. These are search-engine recommendations, not evidence that AI answer products retain citations for the same period.

Track each type of evidence separately. Server logs show requests your servers received. Analytics shows the visits it recorded, and search tools report activity in search. Repeated checks show which citations appeared for specific questions in specific AI products. None alone proves that the move preserved the same visibility or business impact. If an important outside site links to an old URL, ask its owner to update the link when practical, while keeping the redirect for links that remain unchanged.

Key Takeaways

• Inventory cited URLs before launch: Include articles, older URL formats, PDFs, and images that receive links or visits.
• Map each old URL to a relevant destination: Use a permanent redirect to the closest replacement, and return a real error when no equivalent exists.
• Preserve the cited material: Keep the answer and supporting files available so the new destination fulfills the old reference.
• Keep redirects and check outcomes separately: Retain redirects long term, then distinguish requests, search activity, and observed AI citations.
• Treat a working redirect as a route, not a guarantee: It does not prove crawler access, indexing, updated citations, or retained business impact.

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com