Honestly, the whole "AI handles enterprise site migrations" thing gets oversold. I've been through a few major rebrands and mergers as well, and the reality is brutal: for the first couple of months, you're essentially starting from scratch on the search side. The machines have to re-learn everything. That said, if you've done the technical groundwork properly - proper redirect mapping, semantic markup, clean site architecture - you'll often bounce back within 2-3 months. I've seen it happen twice last year with clients who merged, they were barely off the board after the first 48 hours.
Would I trust an AI to automate the entire remapping process? Sort of. I'd spot-check a solid sample size of redirects manually, but a few AI errors won't kill you - same as a few human errors. The bigger risk is volume. AI works well with structured data (site maps, old URL lists, new URL lists). It's just building a "this was here → now it's here" dataset. That's a good use case for automation.
One thing I'd hammer home: don't redirect content that's genuinely dead to a close-but-not-exact match. That creates a credibility gap. The AI (and Google's index) has a cached version of the old page, if it follows a redirect and finds completely different or missing content, you erode trust signals fast. Let it 404 cleanly. If you must redirect, be precise.
Regardless of how good your tooling is, expect a dip - it's unavoidable. Make sure the new site has all the semantic markup, structured data, and tech SEO fundamentals (Screaming Frog audits, GSC coverage reports, Ahrefs site audit) to prove the new information architecture is coherent. The more the site feels like "the same pages, just different URLs", the quicker recovery happens. In my experience, 2-3 months is the upper end before you're back to pre-migration performance, often sooner.
