A site migration can undo months of AI visibility work in the space of one deployment, and most businesses do not find out until a citation that used to show up in ChatGPT or an AI Overview quietly disappears. Redirects, structured data, and crawler access all live in the parts of a site that get rebuilt during a redesign, replatform, or domain change, and none of them survive by default.
Classic SEO has a rough safety net for this. Google recrawls constantly and eventually notices a fix. Answer engines are less forgiving, because some AI crawlers visit far less often and do not always retry a page that failed the first time. A mistake that would have been temporary five years ago can now sit broken for weeks.
Redirects Are the First Casualty
Every URL that used to carry a citation needs a 301 redirect to its new address, and migrations lose this constantly. A URL structure that changes, a CMS switch, or a rebrand that renames slugs all break the link between the old page an engine already trusted and the new page that replaced it.
Without the redirect, an AI crawler hitting the old URL gets a dead end instead of a page. Any citation built at that address does not transfer, it just stops resolving. A full list of old URLs mapped one to one against their new equivalents, checked before launch and not just assumed, is the difference between carrying forward what already worked and starting over from nothing.
Schema and Structured Data Rarely Survive the Rebuild
Structured data is code, and code gets replaced in a migration unless someone specifically carries it forward. Organization schema, Service schema, FAQ markup, and the rest of what a page relies on to state facts clearly to a machine often lives in a template that a new site simply does not include.
This is the same structured data we described building out in our post on schema markup worth shipping. Building it once is only half the job if a platform change three years later quietly strips it back out. A migration checklist should treat schema as content, not as decoration, and verify every type that existed before still exists after.
AI Crawlers Do Not Always Come Back to Check
Some AI crawlers visit a site on a slower, less predictable schedule than Googlebot. A page that returns an error or gets blocked during a migration window may not get revisited for a while, and until it does, it cannot be cited. That gap has real cost during exactly the period a business is trying to relaunch and get found again.
Timing the migration matters more than most teams plan for. Launching cleanly, verifying crawlability immediately, and confirming key pages resolve correctly within hours rather than days shrinks the window where an engine finds nothing where a citation used to be.
robots.txt Gets Rewritten by Default, Not on Purpose
Most platforms generate a new robots.txt automatically when a site moves or relaunches, and that default rarely accounts for which AI crawlers a business actually wants to allow. We covered the specifics of this in our post on which bots to allow and which to block. A migration is exactly the moment that careful configuration gets silently overwritten by a generic template nobody reviewed.
Checking robots.txt after every migration, not just at initial setup, catches this before it costs weeks of invisibility. A single disallow rule copied from a template can block GPTBot or PerplexityBot without a single person deciding that was the goal.
Entity Consistency Breaks When URLs and Domains Change
A domain change or a name change during a rebrand can knock the new site out of sync with everywhere else the business is listed. We wrote about why this matters in our post on entity consistency across surfaces. A directory listing, a Google Business Profile, or a schema reference still pointing at the old domain creates exactly the kind of disagreement that makes an answer engine less confident in any of the sources.
Updating the canonical domain everywhere it is referenced, not just on the site itself, is part of what makes a migration safe instead of merely completed. A new site that looks finished but still disagrees with its own footprint elsewhere is not actually done.
A Migration Needs an AI Visibility Checklist, Not Just a QA Pass
Standard QA checks that pages load and forms submit. It rarely checks that every old URL redirects, that schema types match what existed before, that robots.txt still allows the right crawlers, and that the entity record agrees across every surface that references the new domain. Those four checks are what actually protect AI visibility through a migration, and none of them show up on a typical launch checklist unless someone adds them.
Verifying all four before and immediately after launch, rather than discovering a gap when a citation disappears weeks later, is part of the migration work inside the UM Media SEO / AEO / GEO Foundation, from $7,500. A migration should be the moment a site gets better, not the moment it quietly loses ground it already earned. Structure only. No ranking or citation guarantees.