A migration that preserves rankings can still torch AI citations, because models re-retrieve on their own schedule and judge the new pages fresh. Treat citation continuity as its own workstream: baseline everything before freeze, verify schema and crawler access on launch day, and re-scan at day 14 with rollback triggers already agreed.
When bots spend their visit on parameter junk and duplicate paths, your revenue pages wait days for a recrawl. Prune, canonicalize, and block the waste; fresh content only matters if it gets fetched.
LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1. Miss them and you spot competitors a handicap on every query; hit them and speed disappears as an excuse.
Redirect maps get the headlines, but schema parity, crawler access rules, and entity language drift are where visibility actually dies. Baseline everything before freeze and re-verify on launch day, not launch week.
This exact method ships inside our client engagements at $45/hr, estimate in writing first.
Start a ProjectLCP 2.5s, INP 200ms, CLS 0.1. Miss them and every query starts with a handicap. Here is the fix order that actually moves scores.
When crawlers spend their visit on parameter junk, your revenue pages wait days for a recrawl.
Most high-intent searches now end in an extracted answer, not a click. AEO is the discipline of being that answer.